University of Southern California Dissertations and Theses

Bioinformatics analysis of the anti-cancer potency of NMI on non-small cell lung cancer and its potential mechanism

(USC Thesis Other)

Bioinformatics analysis of the anti-cancer potency of NMI on non-small cell lung cancer and its potential mechanism

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

BIOINFORMATICS ANALYSIS OF THE ANTI-CANCER POTENCY OF NMI ON
NON-SMALL CELL LUNG CANCER AND ITS POTENTIAL MECHANISM

by

Yuxuan Lian

A Thesis Presented to the
FACULTY OF THE USC SCHOOL OF PHARMACY
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
PHARMACEUTICAL SCIENCES

May 2021

Copyright 2021 Yuxuan Lian

ii
Acknowledgements
It is a great pleasure to express the deepest appreciation to my master thesis advisor,
Dr. Jean C. Shih, who gave me the best guidance in the past two years. You have incredible
enthusiasm and a serious attitude toward science and are willing to share your valuable scientific
experience with everyone. You gave me sufficient freedom to study the project that I am interested
in, and always gave helpful advice when I was confused. Besides, you care about my life as well
as study and often convey your kindness and warmth, especially in this special year of Covid-19.
I would like to thank all the staff and students in our lab. To the staff, thank you for your
time and guidance on experimental performances. Also, thank you for sharing your experience and
knowledge of experimental principles. To my labmates, thank you for your easygoing and
friendliness, which makes our cooperation go well and pleasantly. Thanks to your help, we are
making progress together.
I would like to thank my family, thank you for your support on my decision to study abroad,
and your emotional support when I was down. You never lose faith in me, which makes me a
confident person. Your support is my power and my backup.

iii
TABLE OF CONTENTS
Acknowledgements ....................................................................................................................... ii
Lists of Tables ............................................................................................................................... iv
Lists of Figures .............................................................................................................................. v
Abstract ......................................................................................................................................... vi
Introduction ................................................................................................................................... 1
Methods .......................................................................................................................................... 9
Results .......................................................................................................................................... 11
1 The potency of NMI on NSCLC: NCI-60 screen data analysis ............................................. 11
1.1 Five-dose curve of NMI: an overall view of screen data ................................................ 11
1.2 Boxplots of NMI: an overall comparison between cancers ............................................ 12
1.3 Z score activity of NMI: cell lines sensitivity or resistance investigation ...................... 14
1.4 Waterfall plots of NMI: cell lines potency comparisons ................................................ 16
1.5 3D plot of NMI: visualization of cell lines potency comparisons .................................. 18

2 The potency of NMI on NSCLC: efficacy and safety comparisons ...................................... 19
2.1 Heatmap of NMI: efficacy comparisons between NMI and marketed NSCLC drugs ... 19
2.2 Therapeutic index: safety profile of NMI ....................................................................... 23

3 The mechanism of action of NMI on NSCLC ....................................................................... 26
3.1 COMPARE algorithm: pattern similarity between NMI and marketed NSCLC drugs.. 26
3.2 Pearson correlation coefficient: NMI and MAO A-related gene correlation .................. 27
3.3 Pearson correlation coefficient: NMI and NSCLC-related gene correlation .................. 30
Discussion..................................................................................................................................... 41
Conclusion ................................................................................................................................... 47
References .................................................................................................................................... 49
iv
Lists of Tables

Table 1 FDA-approved drugs for NSCLC treatment. .................................................................. 20
Table 2 Pairwise PCCs of NMI with 15 FDA-approved NSCLC drugs by COMPARE, and
their mechanisms. ......................................................................................................................... 27
Table 3 Correlations of 9 genes with NMI, MAO A, their regulation on NSCLC, and their
mechanisms. .................................................................................................................................. 40
Table 4 Top 10 pairwise PCCs of NMI with all marketed FDA-approved drugs by
COMPARE. .................................................................................................................................. 42
Table 5 Top pairwise PCCs of NMI with all unmarketed drugs by COMPARE. ....................... 44

v
Lists of Figures

Figure 1 Chemical structure of NMI. ............................................................................................. 2
Figure 2 RNA-seq expression overview of MAO A on normal tissues......................................... 4
Figure 3 The overview of MAO A expression. ............................................................................. 6
Figure 4 Five-dose response curves of NMI on 9 types of cancer. .............................................. 11
Figure 5 Boxplots of 9 types of cancers with NMI treatment. ..................................................... 13
Figure 6 Z score activity for NMI against 59 cancer cell lines. ................................................... 15
Figure 7 Waterfall plots of 59 cancer cell lines treated with NMI. ............................................. 17
Figure 8 3D scatter plot of 59 cancer cell lines treated with NMI. .............................................. 18
Figure 9 Heatmaps of comparison between NMI and 15 FDA-approved NSCLC drugs. .......... 22
Figure 10 Therapeutic index of NMI and 15 FDA-approved NSCLC drugs for 9 NSCLC
cell lines. ....................................................................................................................................... 25
Figure 11 PCCs of MAO A expression level on 9 NSCLC cell lines. ........................................ 28
Figure 12 Scatter plots of top 10 Pearson’s correlation between NMI (GI50) and gene
expression level on 9 NSCLC cell lines. ...................................................................................... 31
vi
Abstract
According to our previous work, prostate cancer and glioma showed increased expression
levels of monoamine oxidase A (MAO A), and near-infrared (NIR) dye conjugate MAO A
inhibitor (NMI) is effective for the treatment of both cancers. This study investigated the potency
of NMI on another MAO A-increased cancer, non-small cell lung cancer (NSCLC), and its
potential mechanism via bioinformatics analysis.
Based on the NCI-60 screen data, results showed that 6 cell lines reached 100% growth
inhibition and 3 cell lines reached 50% growth inhibition out of 9 NSCLC cell lines at 10 μM NMI.
Among these 9 cell lines, more than half exhibited sensitivity against NMI, the sensitivity was
determined by the Z score of GI50, TGI, and LC50 values. Additionally, NMI showed especially
great potency on 4 out of 9 NSCLC cell lines. Thus NMI has the great potential for NSCLC
treatment. Next, comparisons of the potency between NMI and other FDA-approval NSCLC drugs
showed that NMI outperformed most of the marketed drugs. What’s more, the safety profile
showed that NMI is safe because its therapeutic index on all 9 cell lines were in the safety range.
Finally, the COMPARE algorithm of mechanism similarity showed that NMI may have a unique
mechanism compared with other FDA-approved drugs. The gene correlation of NMI analysis
showed that in addition to MAO A, NMI might regulate several other genes. The interaction of
these genes and MAO A may be the mechanism of NMI function. In summary, this study showed
that NMI is a potent drug for NSCLC, whose mechanism is unique compared to existing drugs. In
addition to MAO A inhibition, there may be more genes involved in the mechanism of NMI, which
warrants further investigation.
Key words: MAO A, NMI, NSCLC, NCI-60 screening, mechanism
1
Introduction
MAO A: a mitochondrial enzyme
Monoamine oxidase A, usually known as MAO A, is a mitochondrial enzyme that
functions as degrading monoamine neurotransmitters and dietary monoamines via oxidative
deamination. Our previous studies have shown the upregulation of MAO A expression level in
prostate cancer and glioma.
1, 2
The possible mechanism of MAO A is, that the overexpression of
MAO A in human prostate cancer cell lines induced epithelial-mesenchymal transition (EMT) via
the production of reactive oxygen species (ROS). ROS inhibited hydroxylase 3 (PHD3), thus
elevated the stabilization of hypoxia-inducible factor 1 α (HIF1α). This process led to the target
genes’ upregulated expression, including vascular endothelial growth factor A (VEGFA), glucose
transporter 1 (GLUT1), and Twist Family BHLH Transcription Factor 1 (TWIST1). Besides, the
interaction of VEGF and neuropilin-1 (NRP-1) activated phosphoinositide 3-kinase (PI3K)/
protein kinase B (AKT) signal transduction pathway, resulting in the nuclear export of
Phosphorylation of Forkhead box protein O1(FOXO1) transcription repressor to the increased
expression of TWIST1, for the promotion of EMT. In terms of mechanism, the ROS generate by
MAO A augmented hypoxic effects, which increased the cancer progression via the tumor
microenvironment. Overall, the upregulated level of MAO A expression promoted ROS
production, EMT, and tumor hypoxia, and resulting in prostate tumorigenesis, progression and
metastasis.
3

NMI: a dual function MAO A inhibitor
NMI, an abbreviation for near-infrared (NIR) dye conjugate MAO A inhibitor, is a novel
MAO A inhibitor synthesized by our lab (Figure 1). As the name implies, NMI consists mainly
of two sections: NIR heptamethine cyanine dye MHI-148 and MAO A inhibitor Clorgyline.
2

Figure 1 Chemical structure of NMI.
The structure was color-coded by conjugation: MAO A inhibitor Clorgyline = green; NIR dye
MHI-148 = pink.
MAO A, monoamine oxidase A; NIR, near-infrared.

Clorgyline is a specific MAO A inhibitor with high affinity. Generally, MAO A inhibitor
has been used as an anti-depressant for decades. They tend to accumulate in the central nervous
system (CNS) and peripheral tissues where there has a high expression of MAO A. To increase
the target specificity to the tumor, NMI was designed to conjugate NIR dye. NIR dye MHI-148
was reported to be absorbed and retained only in cancerous cells of human origin (lung, prostate,
etc.) with no systemic toxicity, not normal cells, thus greatly elevating the specific targeting. MHI-
148 accumulated in the mitochondria and lysosomes of tumor cells, probably through the
mediation of the OATP family proteins, because the active uptake and retention can be blocked by
bromosulfophthalein. Another function of NIR dye is that, when alone or co-stained with the
mitochondrial specific dye MitoTracker Green, NMI can be rapidly tracked by non-invasive
imaging, which makes NMI a dual-function agent, for both diagnosis and therapy for cancers.
1, 4

3
Our previous studies have shown that NMI is effective in treating two MAO A-increased
cancers, prostate cancer and glioma.
Studies on prostate cancer showed that in vitro, NMI suppressed the number of
proliferating prostate cancer cells and the formation of the colonies, then reduced their migration
and invasion. In vivo, NMI delayed the tumor growth, reduced the Serum prostate-specific antigen
(PSA) level, and decreased the tumor weight in subcutaneous prostate tumor xenograft mouse
models, with no accumulation detected in normal tissues.
1

Glioma is a type of tumor that occurs in the brain and spinal cord, brain cancer, and spinal
cord cancer together make up CNS cancer. Studies on glioma showed that in vitro, NMI
significantly inhibited the glioma cell invasion by decreasing the number of invasive cells. In vivo,
NMI increased median survival in the mouse models with intracranial U251R glioma cells
implanted. When combined with Temozolomide, a current therapeutic agent for treating newly
diagnosed Glioblastoma, the median survival time was further extended, which indicated that NMI
can enhance the therapeutic efficacy of Temozolomide. Other in vivo experiments suggested that
NMI reduced the proliferation, invasion, and angiogenesis of glioma, which contributed to the
extended survival, and increased macrophage density, therefore enhancing macrophage
accumulation.
2

Since NMI has been reported effective to inhibit the growth of prostate cancer and glioma,
a reasonable hypothesis is put forward based on these experimental results that NMI is likely to
show effectiveness to other types of cancer, if any, with increased MAO A expression, too. Based
on this hypothesis, the regulation of MAO A expression on various cancers was searched and
compared.

4
NSCLC: a disease with high mortality and high MAO A expression

Figure 2 RNA-seq expression overview of MAO A on normal tissues.
RNA data were collected from three different datasets, respectively: the Human Protein Atlas
(HPA) RNA-seq dataset, the Genotype-Tissue Expression (GTEx) project RNA-seq dataset,
FANTOM5 project CAGE data, and the consensus dataset, which is based on a combination of
these 3 datasets.
5

Color-coded was based on tissue groups, green = Lung.
5
According to the overview of MAO A expression provided by The Human Protein Atlas
(https://www.proteinatlas.org/), MAO A expression varies significantly in different organs of
healthy humans. Among 4 datasets (Consensus dataset, HPA dataset, GTEx dataset, and
FANTOM5 dataset), lung all shows high RNA-seq expression of MAO A (see green bars/arrows
in Figure 2), which could be utilized as a target if lung cancer patients show even higher expression
of MAO A.
5, 6

Lung cancer is one of the most dreadful cancers worldwide due to its enormous numbers
of deaths, which make up almost 25% of deaths caused by all cancer. In 2020, World Health
Organization (WHO) estimated that there were about 2.2 million newly diagnosed cases of lung
cancer, which accounts for approximately 11.4% of the global cancer burden, and 1.79 million
deaths of lung cancer.
7
Based on the American Cancer Society’s estimation for lung cancer
(https://www.cancer.org/), in 2021 the United States would have about 235,760 newly diagnosed
cases of lung cancer (119,100 in men and 116,660 in women), and about 131,880 deaths from lung
cancer (69,410 in men and 62,470 in women).
8
There are two main types of lung cancer, small cell
lung cancer (SCLC) and non-small cell lung cancer. Approximately, 10% to 15% of lung cancer
belongs to SCLC, and 80% to 85% of lung cancer belongs to NSCLC, which can be further divided
into three main subtypes: adenocarcinoma (ADC), squamous cell carcinoma (SCC), and large cell
carcinoma (LCC). Other subtypes of NSCLC, such as adenosquamous carcinoma and sarcomatoid
carcinoma, are much less common. Besides, the uncertain histological subtype is described as
“not otherwise specified” (NOS).
9, 10
Therefore, it is possible that MAO A expression differentiates
in different subtypes of NSCLC.
To investigate the correlation between NSCLC and MAO A, on one hand, Gene Expression
database of Normal and Tumor tissues 2 (GENT2 database, http://gent2.appex.kr/gent2/) was used
6
to compare MAO A expression levels between subtypes of NSCLC. Figure 3A showed that MAO
A expresses in each subtype of NSCLC. Among them, NSCLC (NOS), ADC and SCC had similar
expression levels around value 9.6 (log2 normalized), and LCC showed a bit lower level, around
8.7, which indicated that there was not much difference between subtypes on MAO A distribution
and expression.
11

Figure 3 The overview of MAO A expression.
A Boxplot of MAO A expression comparison by NSCLC subtypes retrieve from Gene Expression
database of Normal and Tumor tissues 2 (GENT2) database.
11

X-axis represented normalized log2 RNA-seq expression, y-axis represented NSCLC subtypes.
B The clinical value of MAO A expression comparison by NSCLC subtypes.
Z score were calculated using the PREdiction of Clinical Outcomes from Genomic Profiles
(PRECOG) database. Comparison results were color-coded by compared types: fold change = red;
Z score = blue. Higher the value, darker the color.
MAO A, monoamine oxidase A; NSCLC, non-small cell lung cancer; ADC, lung adenocarcinoma;
LCC, large cell carcinoma; SCC, squamous cell carcinoma.

On the other hand, there was various evidence indicating that MAO A is upregulated in
NSCLC in recent years (Figure 3B). In 2018, Liu et al. found that NSCLC tissue was shown to
7
have a higher level of protein and mRNA expression levels of MAO A, the relative mRNA-seq
expression in cancer tissue was reported about 2.5-fold higher compared to non-tumor adjacent
tissue.
12
PREdiction of Clinical Outcomes from Genomic Profiles (PRECOG database,
https://precog.stanford.edu/index.php) compared MAO A expression levels of normal tissues with
NSCLC tumor tissues via 4 datasets (GSE19188, GSE11969, GSE1037, Roepman’s reference),
and calculated Z score.
13
Results indicated that LCC expressed slightly increased MAO A mRNA
level (Z score value = 0.24). PRECOG also evaluated MAO A expression level via The Cancer
Genome Atlas (TCGA) Program RNA-seq dataset. Results showed that MAO A was increased in
SCC (Z score value = 1.34). Besides, Cancer RNA-Seq Nexus (CRN database,
http://syslab4.nchu.edu.tw/index.jsp) displayed the RNA-seq expression of MAO A both in lung
adenocarcinoma tumor tissue (0.64 FPKM) and normal tissues (0.28 FPKM) via dataset
GSE52248.
14
Results showed that MAO A expression is up-regulated in tumor tissues than normal
tissues, with 2.29-fold change. Similarly, EMBL-EBI database (https://www.ebi.ac.uk/gxa/home)
also reported a case of increased 1.5 log2-fold change MAO A expression in lung adenocarcinoma,
and 2 cases of increased mRNA expression of MAO A in LCC (log2-fold change = 2; 1.7).
15

Currently, the treatment of NSCLC is based on the stage of this disease. Early-stage
NSCLC (Stage 0/I/II) is usually treated with surgery. The treatment for the advanced stage of
NSCLC (stage IIIA/IIIB/IVA/IVB) is more complex and controversial. Treatments include
chemotherapy, radiation therapy, targeted therapy drugs, immunotherapy, or some of their
combinations. If NSCLC progresses or recurs after treatment, all treatments might be choices for
patients, depending on the situation.
Over the past 10 to 20 years, the discovery and development of targeted therapy and
immunotherapy have unprecedentedly improved the therapeutic effect of NSCLC. Parts of these
8
drugs are now becoming the first-line treatment alone or combined. However, there are still plenty
of questions that existed, for example, the treatments only showed efficacy on specific subtypes of
NSCLC, and the efficacy varied greatly on patients, with just a minority of them benefit a lot. In
this case, an ideal, perfect agent for treating NSCLC does not show up yet, thus more new agents
are welcomed to join this field with massively potential in the future.
8, 9, 16

As NSCLC was found to express an elevated MAO A level in tumor tissues than normal
tissues, further investigation will focus on the efficacy of NMI for the treatment of NSCLC, which
drives the utilization of one cancer-related program named the NCI-60 Human Tumor Cell Lines
Screen.
NCI-60 tumor cell lines screen: an identifier for potential anti-cancer agents
The NCI-60 Human Tumor Cell Lines Screen (NCI-60) is a program generated by the
Developmental Therapeutics Program (DTP), National Cancer Institute (https://dtp.cancer.gov).
So far, NCI-60 has become one of the most widely recognized resources for testing new
compounds available to the global cancer research community. This screen aims to identify and
characterize novel agents (small molecules, proteins, or other) showing potential for cancer
inhibition or treatment. The NCI-60 screen utilized 60 different human tumor cell lines to represent
nine different types of cancer, which include leukemia, NSCLC, colon cancers, CNS cancers,
melanoma, ovarian cancers, renal cancers, prostate cancers, and breast cancers.
17

All screening were tested initially at one-dose in the full NCI 60 cell panel. Only
compounds that exhibited significant growth inhibition and satisfy the threshold inhibition criteria
will progress to the full five-dose assay. Five-Dose screening results were described by three dose-
response parameters, GI50, TGI, and LC50.
17

9
All screen data for compounds accepted were added into DTP in vitro screening database,
where data of most public agents are available to capture. This database generated unique pattern
recognition algorithms (COMPARE) that can characterize whether the mechanism of action of the
tested agent is similar or not to any other standard prototype compounds included in this database
by comparing their biological response patterns.
18

Methods
NCI-60 screening methodology
NMI was initially tested at a one-dose (10
-5
M) assay, then a five-dose assay because
NMI met the threshold inhibition criteria. This thesis analyzes NCI-60 data mainly based on the
data of five-dose assay, thus the five-dose screening methodology will be described. The detailed
description is available at the NCI-60 Screening Methodology website
(https://dtp.cancer.gov/discovery_development/nci-60/methodology.htm).
Seed cells in 96 well microtiter plates (desire densities range from 5,000 to 40,000
cells/well) and incubate the cells after inoculation for 24h before the addition of NM. The NMI
addition was at five concentration levels, with 10-fold or ½ log serial dilutions, ranging from 10
-8

to 10
-4
M. After addition, incubate the plates for 48 h, then stain and read the absorbance. Three
dose-response parameters (GI50, TGI, LC50) were calculated according to seven absorbance
measurements, time zero (Tz), control growth (C), and test growth at the five concentration levels
(Ti). GI50, is NMI concentration required for 50% cell growth inhibition, calculated from [(Ti-
Tz)/(C-Tz)] x 100 = 50; TGI, total growth inhibition, is NMI concentration required for 100 % cell
growth inhibition (0 growth percentage, calculated from Ti = Tz; LC50, 50% lethal concentration,
is NMI concentration required for 50 % cells death, calculated from [(Ti-Tz)/Tz] x 100 = -50. If
10
the effect is not reached or is exceeded, the value for that parameter is expressed as greater or less
than the maximum or minimum concentration.
17

COMPARE algorithm
NMI was specified by using the compound’s NCI accession number (the NSC number),
then the online COMPARE portal (https://dtp.cancer.gov/databases_tools/compare.htm) was
proceeded to rank the entire database. This algorithm calculated the mean cell graph “fingerprints”
or patterns in the order of the similarity of the responses between NMI and other tested compounds.
The similarity of patterns will be expressed quantitatively as a Pearson correlation coefficient
(PCC).
18

Visualization
PerkinElmer ChemDraw Professional, v. 20.0.0.38 was used for visualizing the chemical
structure of NMI.
RStudio v. 1.3.1073 and R v. 3.6.1 were used for boxplot, waterfall plots, 3D plot, heatmap,
and scatter plots.
Statistics
Comparisons between cancers were performed by the Kruskal-Wallis H test. For Figure 5,
with more than 2 groups of cancer, the overall p-value (comparing all groups) was less than 0.05,
therefore pairwise comparisons were made to compare each cancer to the NSCLC. No adjustments
were made for pairwise comparisons. A p-value of less than 0.05 was considered statistically
significantly different.
All correlations were calculated based on Pearson Correlation Coefficient, performed on
RStudio v. 1.3.1073 and R v. 3.6.1, and CellMinerCDB, version 1.2.

11
Results
1 The potency of NMI on NSCLC: NCI-60 screen data analysis
1.1 Five-dose curve of NMI: an overall view of screen data

Figure 4 Five-dose response curves of NMI on 9 types of cancer.
9 cancers were represented by 59 cell lines, and curves were obtained from NCI-60. X-axis
represented the five concentration levels of NMI, 10
-4
M to 10
-8
M. Y-axis represented the
percentage growth, the 100% growth represents the observed growth of cells without treatment.
The 0% growth indicated no observed growth of cells, corresponding to the number of cells at the
start point. The -100% growth indicated all cells killed by the treatment.
GI50, growth inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %.

Figure 4 showed 59 human cancer cell lines, representing 9 types of cancer used by NCI-
60. Five-dose levels of NMI were treated, from 10
-4
M to 10
-8
M. Three dose-response parameters,
GI50, TGI, and LC50 were calculated and utilized for further analysis. GI50 is the NMI
12
concentration required for 50% cell growth inhibition; TGI, total growth inhibition, is the NMI
concentration required for 100 % cell growth inhibition (0 growth percentage); LC50, 50% lethal
concentration, is the NMI concentration required for 50 % cells death.
With 10 μM (10
-5
M) treatment of NMI, overall 48 cell lines reached 100% growth
inhibition, 9 cell lines reached 50% growth inhibition, and 2 remaining cell lines did not reach 50%
growth inhibition. To NSCLC, 6 cell lines reached 100% growth inhibition and 3 cell lines reached
50% growth inhibition out of 9 cell lines.
1.2 Boxplots of NMI: an overall comparison between cancers
To start the analysis, an overall comparison between all types of cancer and NSCLC was
performed on GI50, TGI, and LC50 via boxplots. Figure 5 showed that on GI50, breast cancer,
CNS cancer, leukemia, and melanoma had significant difference with NSCLC (*p < 0.05, **p <
0.01); on TGI, melanoma had significant difference with NSCLC (p < 0.01); on LC50, leukemia
and melanoma have significant difference with NSCLC (p < 0.01). Here, breast and CNS cancer
only showed significant differences on GI50, not TGI and LC50, which means they are not
completely different with NSCLC; leukemia showed significantly greater GI50 median, but worse
TGI and LC50 medians than NSCLC, thus their difference could be complicated but not obvious;
melanoma showed a completely significant difference with NSCLC, but melanoma is not a MAO
A-increased cancer, investigation about melanoma was further discussed later. Putting the results
of the three boxplots together, it can be determined that NSCLC is worth investigating, for NSCLC
shows not much significant difference between different types of cancer, especially within other 2
MAO A-increased cancers, prostate and CNS cancer.
13

Figure 5 Boxplots of 9 types of cancers with NMI treatment.
A Boxplot of GI50. B Boxplot of TGI. C Boxplot of LC50.
X-axis represented 9 types of cancers, y-axis represented -log concentration of GI50, TGI, and
LC50. Boxes were color-coded by cancer types: Breast cancer = gray; CNS cancer = pink; Colon
cancer = light green; Leukemia = orange; Melanoma = blue; NSCLC = red; Ovarian cancer =
purple; Prostate cancer = yellow; Renal cancer = dark green. Data represents the median. The p-
value was calculated by the Kruskal-Wallis H test, *p < 0.05, **p < 0.01.
GI50, growth inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %.

14
NSCLC may not show the best efficacy to the NMI treatment, but it is not bad. In other
words, NMI seems potent on NSCLC and other cancers other than known prostate cancer. Besides,
the boxplots displayed the distribution of each cell line within which cancer it belongs. On GI50,
the boxes are denser-grouped, with the addition of NMI concentration, boxes become loose, which
means the cell lines performed similarly on GI50, but diversely on TGI and LC50. Several
outliners indicate that different cell lines or subtypes of the same cancer type may perform
differently.
1.3 Z score activity of NMI: cell lines sensitivity or resistance investigation
Since boxplots analyzed the data by cancer groups, the next step was to study each cell line
respectively. These 9 NSCLC cell lines covered all 3 main histologic types of NSCLC,
adenocarcinoma, squamous cell carcinoma, and large cell carcinoma. Among them, A549/ATCC,
EKVX, HOP-62, NCI-H522, NCI-H23, and NCI-H322M originated from lung ADC, NCI-H226
originated from lung SCC, and NCI-H460 and HOP-92 originated from lung LCC.
19

To identify cell line’s sensitivity to NMI treatment, z score activity of GI50, TGI, and LC50
provided by the DTP database CellMiner was a useful tool. The z score activity represented the
concentration deviation from the mean via standard deviation. Positive values were plotted to the
right of the vertical line (exceeded the mean), which were regarded as NMI-sensitive; on the
contrary, negative values were plotted to the left of the vertical line (less than the mean), which
were regarded as NMI-resistant. Longer the bar, the more sensitive/resistant the cell line.

15

Figure 6 Z score activity for NMI against 59 cancer cell lines.
A Z score activity of GI50. B Z score activity of TGI. C Z score activity of LC50.
Figure was obtained from CellMiner database. X-axis represented the z scores, y-axis represented
cell lines, and vertical line represented the mean. NMI-sensitive cell lines were plotted to the right
of the vertical line, while NMI-resistant cell lines were plotted to the left of the vertical line. Cell
lines were color-coded by cancer types: Breast cancer = dark blue; CNS cancer = light brown;
Colon cancer = orange; Leukemia = light green; Melanoma = dark green; NSCLC = blue; Ovarian
cancer = dark brown; Prostate cancer = yellow; Renal cancer = red.
GI50, growth inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %.

As shown in Figure 6, there seems to be no distinct sensitivity pattern of NMI. Basically,
each type of cancer has both sensitive and resistant cell lines, and cell lines didn’t show continuous
sensitivity or resistance with the addition of NMI’s treatment. To NSCLC cell lines (highlighted
in blue), there were 2 cell lines representing sensitivity on GI50 (Figure 6A), 5 cell lines on TGI
(Figure 6B), and 4 cell lines on LC50 (Figure 6C). 2 cell lines, HOP-92 and NCI-H522, were
sensitive all the time, whereas others’ sensitivity varied. Regardless of the known efficacy of NMI
on treating prostate cancer, prostate cancer cell lines (highlighted in yellow) were both resistant
16
on GI50 and maintained one sensitive, one resistant on TGI and LC50, which reflected that NMI’s
sensitivity pattern may be discrete for each cell line, and discontinuous in one cell line when
concentration increased. Overall, more than half of NSCLC cell lines exhibited sensitivity against
NMI, especially HOP-92 and NCI-H522, suggesting that NSCLC has the potential to be treated
by NMI. Besides, the 2 most sensitive cell lines belong to ADC and LCC subtypes, and the SCC
cell line was sensitive mostly, indicating that NMI might be capable of treating different NSCLC
subtypes.
1.4 Waterfall plots of NMI: cell lines potency comparisons
Next, the difference between specific cell lines was studied using the waterfall plot. Within
safety ranges, drug potency and dose-response parameters (GI50, TGI, and LC50) are inversely
related, for example, a more potent drug requires a lower concentration of GI50/TGI/LC50. By
reordering cell lines in the order of their potency, the most potent cell line was listed at the top of
the plot, and the following potency decreased progressively. To investigate the drug potency on
NSCLC, prostate cancer was exploited as a comparing standard because it is a cancer known to be
potent to NMI treatment.
When compared by GI50 (Figure 7A), 8 out of 9 NSCLC cell lines showed higher or
equivalent potency than prostate cancer cell lines, PC-3 and DU-145. Among them, 3 cell lines,
NCI-H522, HOP-92, and NCI-H226, had better/equal potency than whole prostate cancer cell lines.
When compared by TGI (Figure 7B), 7 out of 9 NSCLC cell lines showed higher or
equivalent potency than prostate cancer cell lines. Among them, 2 cell lines, NCI-H522 and HOP-
92, had a better potency than whole prostate cancer cell lines.
17
When compared by LC50 (Figure 7C), 4 out of 9 NSCLC cell lines showed higher potency
than prostate cancer cell lines. Among them, 2 cell lines, NCI-H522 and HOP-92, have better/equal
potency than whole prostate cancer cell lines.

Figure 7 Waterfall plots of 59 cancer cell lines treated with NMI.
A Waterfall plot of GI50. B Waterfall plot of TGI. 2 blank bars represented no available data for
these 2 cell lines. C Waterfall plot of LC50.
Waterfall plots were ordered by concentration values, from the most potent concentration (lowest,
at the top), to the least potent concentration (highest, at the bottom). X-axis represented -log
concentration in GI50, TGI, and LC50, y-axis represented 59 cell lines. Cell lines were color-
coded by cancer types: Breast cancer = gray; CNS cancer = pink; Colon cancer = light green;
Leukemia = orange; Melanoma = blue; NSCLC = red; Ovarian cancer = purple; Prostate cancer =
yellow; Renal cancer = dark green.
GI50, growth inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %.

18
1.5 3D plot of NMI: visualization of cell lines potency comparisons
The cell lines’ potency could be better visualized by correlating into one graph, R was used
to draw a three-dimension scatter plot of GI50, TGI, and LC50 (Figure 8).

Figure 8 3D scatter plot of 59 cancer cell lines treated with NMI.
X-axis represented -log concentration in GI50, y-axis represented -log concentration in TGI, and
z-axis represented -log concentration in LC50. Origin of coordinates was on the bottom right. Cell
lines were color-coded by cancer types: NSCLC = red; Ovarian cancer = purple; Prostate cancer
= yellow; Other cancers = gray.
GI50, growth inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %.

The origin of coordinates was shown on the bottom right of the figure, which indicated the
least potent position. Cell lines with higher potency to NMI’s treatment meant to be plotted further
from the origin, thus the top left corner is the most potent position. From the 3D plot, it was clear
that between 2 prostate cancer cell lines (highlighted in yellow), PC-3 belonged to one of the most
potent cell lines with the treatment of NMI, and DU-145 showed middle potency among all cell
19
lines. By comparison, 3 NSCLC cell lines (highlighted in red), NCI-H522, HOP-92, and NCI-
H226 exhibited their potency because they were not only better than or equivalent with PC-3, but
also in the leading position among 59 cell lines. Although HOP-62 was not as potent as these three
cell lines, it was more potent than DU-145, and also a middle-potent cell line.
Summarily speaking, 4 out of 9 NSCLC cell lines (HOP-62, HOP-92, NCI-H226, NCI-
H522) showed great potency compared to prostate cancer cell lines. Specifically, 2 cell lines, NCI-
H522 and HOP-92 outperformed because they exhibited higher potency on all GI50, TGI, and
LC50 than the best-performed cell line of prostate cancer, PC-3. Therefore, NMI is potent on
NSCLC cell lines, and it showed a more potent treatment to the above 4 cell lines than the
remaining 5 cell lines of NSCLC, which suggested the fact that the differences between cell lines
truly exist. In terms of subtypes, each subtype had its representative outstanding cell lines, lung
ADC (HOP-62, NCI-H522), lung SCC (NCI-H226), and lung LCC (HOP-92). These results were
consistent with the z score sensitivity study, implying that there is no specific subtype of NSCLC
that NMI treats for, in contrast, NMI is probably efficient for the whole NSCLC disease.
2 The potency of NMI on NSCLC: efficacy and safety comparisons
2.1 Heatmap of NMI: efficacy comparisons between NMI and marketed NSCLC drugs
Since NMI has the potential to treat NSCLC, further study focused on how potent NMI is.
Although NMI seems potent based on its NCI-60 screening data, the potency of NMI could be just
relatively high to itself, while limited compared to other true NSCLC treating drugs. In this case,
marketed drugs approved by the United States Food and Drug Administration (FDA) can be used
to compare with NMI. Analyzing the NCI-60 data of other FDA-approved NSCLC drugs were
supposed to provide an insight for understanding the ability of NMI.

20
Table 1 FDA-approved drugs for NSCLC treatment.
Drug Name NCI-60 Data Treatment
Alectinib Completed ALK-positive metastatic NSCLC
Carboplatin Completed Chemotherapeutic treatment for NSCLC
Crizotinib Completed ALK or ROS1-positive metastatic NSCLC
Docetaxel Completed Locally advanced or metastatic NSCLC after failure of prior platinum-based chemotherapy
Doxorubicin Completed Chemotherapeutic treatment for NSCLC
Erlotinib Completed EGFR gene mutated metastatic NSCLC
Everolimus Completed Unresectable, locally advanced, or metastatic NSCLC.
Gefitinib Completed EGFR gene mutated metastatic NSCLC; in combination with Everolimus
Lorlatinib Completed ALK-positive metastatic NSCLC
Osimertinib Completed EGFR gene mutated metastatic NSCLC
Paclitaxel Completed Chemotherapeutic treatment for NSCLC
Pemetrexed Completed Locally advanced or metastatic, non-squamous NSCLC
Selpercatinib Completed RET gene mutated metastatic NSCLC
Trametinib Completed In combination with dabrafenib for BRAF V600E mutated metastatic NSCLC
Vinorelbine Completed Alone or in combination with Cisplatin for locally advanced or metastatic NSCLC
Afatinib Incomplete

Brigatinib Incomplete

Ceritinib Incomplete

Dabrafenib Incomplete

Dacomitinib Incomplete

Entrectinib Incomplete

Gemcitabine Incomplete

Methotrexate Incomplete

Atezolizumab No Data

Bevacizumab No Data

Capmatinib No Data

Duralumab No Data

Ipilimumab No Data

Necitumumab No Data

Nivolumab No Data

Pembrolizumab No Data

Pralsetinib No Data

Ramucirumab No Data

Tepotinib No Data

NSCLC, non-small cell lung cancer.

So far, 34 kinds of NSCLC drugs got approval from FDA, among which NCI-60 screening
data were available for 23 drugs and unavailable for 11 drugs. Within 23 available screening drug
data, 15 of them were completed thus utilized for future comparison; 8 were taken out because
data for some cell lines were missing, or the five-dose concentration range was different (not 10
-8

21
to 10
-4
M). Table 1 listed a total of 34 FDA-approved NSCLC drugs, and the treatment of 15 drugs
used for comparison.
Comparisons were initially made with three parameters, GI50, TGI, and LC50, respectively,
then with cumulative score via min-max normalization. The min-max normalization was
performed because GI50, TGI, and LC50 were dose-responded and correlated parameters, the
normalized cumulative score combined them, and provided an all in all conclusion of separated
comparisons.
By min-max normalization, GI50, TGI, and LC50 were normalized, and all scaled into a
range in [0, 1]. The formula is as follows
20
:
𝑥 ′ =
𝑥 − 𝑚𝑖𝑛 (𝑥 )
𝑚𝑎𝑥 (𝑥 ) − 𝑚𝑖𝑛 (𝑥 )

X’ represented the normalized value, x represented the original value, min(x) and max(x)
were the minimum and maximum original values of the whole comparison set. The final score was
scaled in [0,1], 0 represented the score for the most potent drug, and 1 represented the least potent.
For example, to calculate GI50, x was the original GI50 concentration, min(x) was the minimum
GI50 concentration, and max(x) was the maximum GI50 concentration among 15 drugs. After
separated normalization, the final cumulative score was calculated by addition of the three
normalized values, ranging in [0,3], 0 represented score for the most potent drug, and 3 represented
the least potent.
The results were visualized via heatmaps. The comparisons of NMI with 15 FDA-approved
NSCLC drugs on 9 NSCLC cell lines on original GI50, TGI, and LC50 concentrations were shown
in Figures 9A, 9B, and 9C. Since each drug was screened via the same five-dose concentrations,
10
-8
to 10
-4
M, the visualization limits were all set from 0 to 100 μM. The concentration differences
were shown by color variations. The darkest color represented 0 μM, and the lightest color
22
represented 100 μM. Thus, drugs with more potency (lower concentration) would have darker
color represented.

Figure 9 Heatmaps of comparison between NMI and 15 FDA-approved NSCLC drugs.
A Heatmap of GI50 comparison. B Heatmap of TGI comparison. C Heatmap of LC50 comparison.
D Heatmap of cumulative score comparison.
X-axis represented drugs; y-axis represented NSCLC cell lines. Concentrations of GI50, TGI,
LC50 (0 μM to 100 μM), and cumulative score (0.0 to 3.0) were color-coded by values: smaller
the value, darker the color.
FDA, the United States Food and Drug Administration; GI50, growth inhibition of 50 %; TGI,
total growth inhibition; LC50, lethal concentration of 50 %.

Based on the GI50 heatmap, NMI (showed in the first column) inhibited 50% growth of 9
NSCLC cell lines with 1-5 μM treatment, which is absolutely a good GI50 concentration range.
Among all drugs compared, the potency of NMI was relatively in the leading position.
Based on the TGI heatmap, NMI (showed in the first column) inhibited 100% growth of 9
NSCLC cell lines with 1-20 μM treatment, which is still a great performance to TGI. Among all
drugs compared, NMI was one of the leading potent drugs.
23
Based on the LC50 heatmap, NMI (showed in the first column) killed 50% of 7 NSCLC
cell lines with 5-50 μM treatment, whereas 2 cell lines with beyond 100 μM treatment. Among all
drugs compared, NMI was definitely a potent drug, for many drugs need beyond 100 μM to reach
LC50.
The results showed that NMI outperformed on each parameter comparison. Some drugs,
such as Vinorelbine, showed great potency on GI50 (all 0-1 μM), but poor performance on TGI
and LC50. Therefore, normalized cumulative score offered a comprehensive analysis. From
Figure 9D, NMI got the best cumulative score for 5 cell lines (HOP-62, HOP-92, NCI-H322M,
NCI-H460, NCI-H522), and the second-best for the rest 4 cell lines. In summary, all cell lines
performed well, which means that the treating efficacy of NMI will not be affected much by
subtypes.
By all these comparisons, NMI exhibited better potency than most of the marketed FDA-
approved NSCLC drugs, indicating that NMI is a potential anti-cancer drug for NSCLC. These
outcomes corresponded with the previous conclusion, in a general way. Previous potency
investigation indicated that 4 NSCLC cell lines within 59 human cancer cell lines showed high
potency against NMI, and they represented all NSCLC subtypes; above comparisons suggested
that 5 NSCLC cell lines (those 4 NSCLC cell lines all included) still outperformed with NMI’s
treatment by contrast with other NSCLC drugs’ treatments, covering each subtype of NSCLC. The
correspondence illustrated that NMI is not only potent on NSCLC treatment under its own
screening result of 9-cancer, but also potent compared with other drugs.
2.2 Therapeutic index: safety profile of NMI
Albeit a more potent drug requires a lower concentration of GI50/TGI/LC50, there ought
to be a boundary indicating the safety of these drugs. GI50 signifies the growth inhibitory power,
24
TGI signifies a cytostatic effect, and LC50 signifies a cytotoxic effect of drugs, for anti-cancer
agents, cytotoxicity could be a double-edged sword because, on one hand, it is important in killing
tumor cells that grow rapidly, while on the other hand, it may damage normal cells too, resulting
in adverse effects to the human body.
To verify the toxicity of NMI, the safety profile was measured by Therapeutic Index (TI).
TI refers to the ratio of the dose needed for causing adverse effects to the dose needed for the
desired pharmacological effect. Here, GI50 and LC50 were used to calculate TI. The formula is as
follows
21
:
𝑇 ℎ𝑒𝑟𝑎𝑝𝑒𝑢𝑡𝑖𝑐 𝐼𝑛𝑑𝑒𝑥 =
𝐿𝐶 50
𝐺𝐼 50

In general, a drug prefers a high TI to demonstrate its safety and efficacy profile. According
to FDA, drugs will be classified as “Narrow Therapeutic Index (NTI)” if there was “a maximum
of a 2-fold difference between minimum effective and minimum toxic dose or maximum
recommended therapeutic dose”. In other words, TI ≤ 3 is a reasonable boundary to define NTI
drugs.
22

Figure 10 listed TIs of 9 NSCLC cell lines for NMI and 15 FDA-approved NSCLC drugs.
Each dot represented one cell line. This figure plotted the majority of all cell lines, the y-axis limit
was set from 0 to 30, cell lines with higher TI were not shown in the figure (the highest TI was
beyond 18,000). The red horizontal line was the boundary line where TI = 3, so cell lines below
the line will be considered as NTI, and cell lines above will be generally safe. Higher the TI, safer
the treatment of the agent.
As results showed, TIs of NMI were all fallen into the safety range (highlighted in red).
Among them, 3 TIs were in (5,10), and the rest 6 were greater than 10, which could be considered
a very good TI, suggesting that NMI is adequately safe to treat NSCLC. From the whole picture,
25
TI for NMI is relatively similar to others and even better than a few drugs, which implied that NMI
is qualified and competitive. However, a drug that has a narrow TI does not always mean it is bad
and useless, it depends. Take Carboplatin as an example (highlighted in yellow), its TI was
relatively low (=1), but in terms of mechanism of action, Carboplatin is a chemotherapy drug, and
a highly cytotoxic activity could exactly be its way to treat. While other 2 chemotherapy drugs,
Doxorubicin and Paclitaxel, did not have similar safety profiles with Carboplatin. This is an
interesting situation that is worth digging deeper into via investigating the mechanism of these
drugs.

Figure 10 Therapeutic index of NMI and 15 FDA-approved NSCLC drugs for 9 NSCLC cell
lines.
X-axis represented drugs, y-axis represented TI. Red line represented TI = 3. TIs were color-
coded by drugs: NMI = red; Alectinib = dark green; Carboplatin = yellow; Crizotinib = light
purple; Docetaxel = coral; Erlotinib = light blue; Gefitinib = orange; Lorlatinib = light green;
Osimertinib = pink; Pemetrexed = dark blue; Selpercatinib = gray; Trametinib = dark purple. TIs
for Doxorubicin, Everolimus, Paclitaxel and Vinorelbine did not shown because they were
beyond the y-axis limits.
TI, therapeutic index; NSCLC, non-small cell lung cancer.

26
3 The mechanism of action of NMI on NSCLC
3.1 COMPARE algorithm: pattern similarity between NMI and marketed NSCLC drugs
In terms of NMI’s outstanding performance compared with other FDA-approved NSCLC
drugs, NMI probably shared a similar mechanism of action with one or several of these drugs. The
COMPARE algorithm generated by DTP was able to compare and recognize the treating pattern
similarity via the Pearson correlation coefficient, therefore probably explaining the potency.
The pairwise Pearson correlation coefficient was utilized to calculate the correlation of
pattern between NMI and other compounds in the database, compounds with high correlation
coefficient were thought to have a similar pattern with NMI. PCC was scaled in [-1,1], a PCC of
1.0 represented a highly perfect similarity, a PCC of -1.0 represented a highly perfect mirror
similarity, whereas a PCC of 0 represented there was no similarity between the two patterns.
According to DTP, a PCC < 0.7 was regarded as no correlation, and a PCC > 0.8 or PCC < -0.8
was regarded as very strong correlation.
23

As shown in Table 2, the highest PCC of GI50 is 0.59 (Vinorelbine), the highest PCC of
TGI is 0.37 (Doxorubicin), and the highest PCC of LC50 is 0.51 (Everolimus), which are all
considered moderate or weak correlation. Besides, the highest PCCs of the three parameters belong
to different drugs, in other words, none of these drugs showed a similar treating pattern to NMI.
So the mechanism of NMI’s potency on NSCLC is different from these 15 NSCLC drugs (see
mechanism), implying that NMI is likely to have a unique mechanism to treat NSCLC.

27
Table 2 Pairwise PCCs of NMI with 15 FDA-approved NSCLC drugs by COMPARE, and their
mechanisms.
FDA-approved NSCLC Drugs
Drugs PCC-GI50 PCC-TGI PCC-LC50 Mechanism of Action
Alectinib 0.23 0.21 0.15 An ALK and RET inhibitor
Carboplatin / / / An alkylating agent
Crizotinib 0.44 0.34 0.45 An ALK and ROS1 inhibitor
Docetaxel 0.26 0.19 0.15 A taxoid
Doxorubicin 0.58 0.37 0.45 An anthracycline antibiotic
Erlotinib -0.27 0.09 0.03 An EGFR tyrosine kinase inhibitor
Everolimus 0.02 0.13 0.51 A mTOR kinase inhibitor
Gefitinib -0.04 0.03 0.22 An EGFR tyrosine kinase inhibitor
Lorlatinib 0.11 0.01 / An ALK and ROS1 inhibitor
Osimertinib -0.15 0.13 0.48 An EGFR tyrosine kinase inhibitor
Paclitaxel 0.38 0.24 0.44 A taxane; a plant alkaloid
Pemetrexed 0.03 / / An antimetabolite
Selpercatinib 0.04 0.14 / A RET tyrosine kinase inhibitor
Trametinib -0.15 0.26 0.32 A MEK kinase inhibitor
Vinorelbine 0.59 0.16 0.24 A plant alkaloid
PCC ranged from -1 to 1. A PCC of 1.0 represented a perfect correlation, a PCC of -1.0 represented
a perfect mirror correlation, whereas a PCC of 0 represented there was no correlation between the
two patterns. |PCC| > 0.8 was regarded as strong correlation. Five-dose concentrations for drugs
ranged all from 10
-8
to 10
-4
M.
PCC, Pearson Correlation Coefficient; NSCLC, non-small cell lung cancer; GI50, growth
inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %; ALK,
anaplastic lymphoma kinase; RET, rearranged during transfection; EGFR, Epidermal growth
factor receptor; MEK, MAPK/ERK kinase.

3.2 Pearson correlation coefficient: NMI and MAO A-related gene correlation
To further study the mechanism of action of NMI on NSCLC, gene correlation was
calculated.
First of all, the correlation between NMI and MAO A, the gene its design derived from,
was calculated by Pearson correlation coefficient (R). As previously described, GI50 signifies the
growth inhibitory power, so the correlation could tell the antiproliferative efficacy of NMI to the
expression level of MAO A. However, the calculation showed that the correlation between GI50
of NMI and MAO A on NSCLC cell lines is - 0.086, which is way smaller than 0.7 (Figure 11A).
28

Figure 11 PCCs of MAO A expression level on 9 NSCLC cell lines.
A Scatter plot of correlation between NMI (GI50) and MAO A expression level. X-axis
represented log concentration in GI50, y-axis represented RNA-seq gene expression levels, plots
represented 9 NSCLC cell lines.
B Heatmap of correlation of expression levels between top MAO A-related genes and MAO A,
and NMI respectively. R and the p-value were calculated by PCC, R ranged from -1 to 1. A |R| of
1.0 represented a perfect correlation, an R of 0 represented there was no correlation. |PCC| > 0.7
was regarded as strong correlation. Heatmap was color-coded by correlation types: positive
correlation = red; negative correlation = green. Higher the correlation, the darker the color.
MAO A, monoamine oxidase A; NSCLC, non-small cell lung cancer; GI50, growth inhibition of
50 %.

As a MAO A inhibitor, NMI was proved efficient to inhibit MAO A expression on prostate
cancer, while this result put up with 2 hypotheses: 1) the expression of MAO A was influenced by
other genes and therefore weakening its correlation with NMI; 2) NMI might play a role in
regulating other genes, in addition to inhibit MAO A.
Assuming the first hypothesis was correct, this possibility led to the investigation of MAO
A’s expression correlation with MAO A-related genes. According to the literature search, genes
that were claimed to be the most related to MAO A expression (n = 120), were chosen to calculate
29
their correlation coefficient.
24
The calculation was done using CellMinerCDB, an online analysis
database generated by DTP. Since the correlation calculation was only focused on NSCLC, the
results might be inconsistent with the literature search because general correlation treated gene
expression as a whole (whole body or all diseases), not so specific on one cancer. In general, a R
of 0 - 0.3 (0 to -0.3) means there is no correlation; a R of 0.3 - 0.5 (-0.3 to -0.5) means low
correlation; a R of 0.5 - 0.7 (-0.5 to -0.7) means moderate correlation; a R of 0.7 - 0.9 (-0.7 to -0.9)
means high correlation; and a R of 0.9 – 1.0 (-0.9 to -1.0) means very high correlation.
25

Figure 11B (the first row) showed the highest 11 gene correlation with MAO A (|R| > 0.5,
moderate correlation or higher), with 8 genes positively correlated and 3 genes negatively
correlated with MAO A expression level. Among them, FAM133A, TRIM55, SEMA3E, and
RCSD1 showed high positive correlations with MAO A. Positive correlation of gene expression
indicates that these 2 genes, MAO A and another, shared the same tendency on NSCLC cell lines
(both increased or decreased) with the influence of each other. On the contrary, a negative
correlation means that MAO A and another gene impeded each other’s expression, when one gene
increased, the other would decrease, and vice versa.
Then, the correlations between these 11 genes and NMI on NSCLC were studied (Figure
11B, the second row). Results showed that among 8 positive-related genes, correlations of 7 genes
to NMI-GI50 on NSCLC cell lines were all below the absolute value of 0.3, which suggested that
they are not correlated to NMI’s treatment on NSCLC. The correlation with the rest one,
ATP6V0D1, was negatively low. As for the 3 negative-related genes, RHPN2 showed a positively
high correlation; IQCA1 and CD37 showed negatively moderate and near-moderate correlations
reversely. The differentiation happening on the 3 MAO A negative-related genes is significant and
interesting, further investigation will be made in the following.
30
To verify the second hypothesis, NMI might play a role in regulating other genes in
addition to inhibit MAO A for NSCLC treatment, the following analysis turned to another studying
direction, to investigate NSCLC-related genes.
3.3 Pearson correlation coefficient: NMI and NSCLC-related gene correlation
Similarly, based on the literature search, the expression levels of genes (n = 89) that were
determined up or down-regulated significantly on multiple subtypes of NSCLC were utilized to
study their correlation with GI50 of NMI.
26, 27

Figure 12 listed the 10 most positive and 10 most negative correlation coefficients of genes
expression level and NMI (GI50) on 9 NSCLC cell lines (3 MAO A negative-related genes were
also included, the remaining 17 were NSCLC-related). In each plot, the x-axis represented the
GI50 concentration of NMI, and the y-axis represented gene expression. Among the top 10
negative correlation genes (plots highlighted in red), CDHR2 and SCL14A2 showed high
correlations with NMI (R < -0.7), and 6 showed moderate correlations, including MAO A-related
gene IQCA1). From the top 10 positive correlation genes (plots highlighted in blue), TMC5 and
RHPN2 showed high correlations (MAO A-related gene included), and 3 showed moderate
correlations.

31

Figure 12 Scatter plots of top 10 Pearson’s correlation between NMI (GI50) and gene expression
level on 9 NSCLC cell lines.
A CDHR2 expression correlation. B SLC14A2 expression correlation. C IQCA1 expression
correlation. D CRNN expression correlation. E SCGN expression correlation. F SOX10
expression correlation. G TP53AIP1 expression correlation. H ST8SIA3 expression correlation. I
CD37 expression correlation. J TOP2A expression correlation. K TMC5 expression correlation.
L RHPN2 expression correlation. M LGALS4 expression correlation. N TM4SF4 expression
correlation. O PHGR1 expression correlation. P MUC5B expression correlation. Q SPP1
expression correlation. R MSMB expression correlation. S GSTA1 expression correlation. T
MSLN expression correlation.
A-J, top 10 negative correlation; K-T, top 10 positive correlation. X-axis represented log
concentration in GI50, y-axis represented RNA-seq gene expression levels. Scatter plots were
color-coded by correlation types: negative correlation = red; positive correlation = blue. R and the
p-value were calculated by Pearson Correlation Coefficient, R ranged from -1 to 1. A |R| of 1.0
represented a perfect correlation, an R of 0 represented there was no correlation between NMI and
gene. |PCC| > 0.7 was regarded as strong correlation.
GI50, growth inhibition of 50 %; NSCLC, non-small cell lung cancer.

32

Figure 12 Cont.
33

Figure 12 Cont.
34
It is known that the lower GI50 concentration means the drug is more potent, so the
correlation between genes and GI50 will indicate the sensitivity or resistance of the gene. A
positive correlation implies that the upregulation of the gene leads to increased resistance to the
drug because increased GI50 is a bad sign of the drug’s efficacy. On the contrary, a negative
correlation indicates that the upregulated gene expression level decreases the drug’s GI50,
resulting in better treatment efficacy.
28

To further analyze the relationship between NMI and these genes, and try to find out the
mechanism of NMI on NSCLC, genes with correlation coefficients greater than 0.6 or less than
-0.6 were selected to analyze.
3.3.1 CDHR2
Cadherin Related Family Member 2 is a protein-coding gene belonging to the
protocadherin family, whose function is to encode non-classical cadherin, a calcium-dependent
cell-cell adhesion molecule. Figure 12A was the highest negative correlation between GI50 of
NMI with all genes calculated (R = -0.74), indicating that with higher expression of CDHR2, NMI
tended to show a better, lower-dose needed GI50. Besides, CDHR2 was reported to be upregulated
in lung ADC.
27
Combining correlation and regulation, it can be inferred that the increased CDHR2
helps NMI, making it more efficient on NSCLC treatment, or NMI may directly inhibit CDHR2.
However, the R between CDHR2 and MAO A was -0.07, which was considered no correlation
between the 2 genes, thus, NMI has a different mechanism related to CDHR2 on NSCLC rather
than MAO A.
3.3.2 SLC14A2
Solute Carrier Family 14 Member 2 is a protein-coding gene. The related pathways are the
transportation of glucose and other sugars, bile salts and organic acids, metal ions, and amine
35
compounds. SLC14A2 was found increased expression in ADC.
27
Figure 12B showed a high
correlation of SLC14A2 and NMI negatively, which offers a possibility that the upregulation of
SLC14A2 in lung tumor tissue than in normal tissue may promote the efficacy of NMI for the
concentration needed for GI50 is less. In the meantime, SLC14A2 and MAO A showed a weak
but slightly positive correlation (R = 0.32), which means when the expression of SLC14A2
increased in NSCLC, MAO A increased, too. It is possible that NMI directly inhibits SLC14A2
because SLC14A2 was correlated to MAO A; if not, NMI may show its efficacy by inhibition of
the increased MAO A induced by SLC14A2. Unfortunately, current investigations about pathways
of SLC14A2 and NSCLC or other cancers are limited, the mechanism of SLC14A2 is still
unknown, further attention could be focused on this gene.
29

3.3.3 IQCA1
IQ Motif Containing With AAA Domain 1, a protein-coding gene, belongs to the ATPases
Associated with diverse cellular Activities (AAA) superfamily, which participates in a large
number of cellular processes and contains the ATPase module consisting of an alpha-beta-alpha
core domain and the Walker A and B motifs of the P-loop NTPases. In Figure 12C, IQCA1 was
highly negatively correlated with NMI (R = -0.68), and this gene was negatively correlated with
MAO A, too (R = -0.52). Besides, IQCA1 was reported downregulated in NSCLC. Based on the
gene-GI50 correlation, NMI will perform better when IQCA1 expression is higher, however,
NSCLC downregulated IQCA1, and therefore probably diminishing NMI’s efficacy. One possible
guess is that since IQCA1 is negatively related to MAO A, its downregulation leads to MAO A’s
upregulation, and NMI can still work eventually as MAO A inhibitor. A study about triple-negative
breast cancer reported that IQCA1 had a positive correlation to lung metastasis, and inactivating
mutations of IQCA1 might represent a route of metastatic lung dissemination
30
, whereas the
36
mechanism of IQCA1 needs further verification. Combine this study and our results, IQCA1’s
downregulation was related to the formation of NSCLC, which provides a new thought that
upregulating the expression of IQCA1 in NSCLC may elevate the efficacy of NMI.
3.3.4 CRNN
Cornulin is a tumor-related protein-coding gene, also known as squamous epithelial heat
shock protein 53, it might play a role in the mucosal/epithelial immune response and epidermal
differentiation. CRNN was increased in SCC
27
, and CRNN was negatively correlated with NMI
(Figure 12D, R = -0.65), which indicated that the existence of CRNN elevated the NMI’s
sensitivity. As for the mechanism, there is a saying that the upregulation of CRNN increased cell
proliferation through inducing cyclin D1, increased the phosphorylation and activation of
phosphoinositide 3-kinase and Akt.
31
If this mechanism also happens on NSCLC, then NMI is
likely to decrease the tumor cell proliferation by the inactivation of the PI3K and AKT pathways,
similar to the known mechanism of NMI on prostate cancer.
3.3.5 SCGN
Secretagogin, an EF-Hand Calcium Binding Protein, is thought to be involved in KCL-
stimulated calcium flux and cell proliferation. SCGN was found overexpression in SCC
27
, and
negative correlation with GI50 of NMI based on Figure 12E (R = -0.64). Recently, SCGN has
been reported to play a crucial role in cell apoptosis, receptor signaling, and differentiation. Yifeng
et al. found that the overexpression of SCGN in small cell lung cancer inhibited cell apoptosis and
speeded the cell cycle of the G0/G1 phase. Besides, miR-494 directly targeted to SCGN, and
upregulation of miR-494 sensitized the cancer cell line to drug treatment.
32
If the upregulation of
SCGN in NSCLC functioned the same with SCGN in SCLC, then NMI probably also regarded
SCGN as a target, to increase cell apoptosis and arrest cell proliferation by inhibition of SCGN.
37
3.3.6 SOX10
SRY-Box Transcription Factor 10 is a protein-encoding gene that encodes a member of the
SOX (SRY-related HMG-box) family of transcription factors involved in the regulation of
embryonic development and the determination of cell fate. The expression level of SOX10 was
increased in SCC
27
, with the negative correlation of SOX10 and NMI (Figure 12F, R = -0.64), the
increased SOX10 promoted the sensitivity of NMI on NSCLC. However, the mechanism was
different to speculate, for SOX10 has multiple functions in different tumors, its potential roles as
tumor suppressors or promoters depend on tumor status and types. It is a tumor suppressor for
digestive cancers, via suppression of the Wnt/β-catenin pathway, whereas it is a tumor promoter
for melanoma, whose upregulation increased tumor progression.
33, 34
Here, the only thing that can
be told was that NMI does have a somewhat correlation with SOX10, while which role they played
for the treatment is still unclear.
3.3.7 TP53AIP1
Tumor Protein P53 Regulated Apoptosis Inducing Protein 1 encodes a protein that is
localized to the mitochondrion. and it is thought to play an important role in mediating p53-
dependent apoptosis. It was reported that a decrease of TP53AIP1 in the lung was considered to
be a predictor of poor prognosis in NSCLC patients. The correlation between TP53AIP1 and NMI
was negative (Figure 12G, R = -0.63), and the expression of TP53AIP1 was increased in SCC.
27

According to a recent study, miR‑505‑5p was upregulated in lung ADC and acted on promoting
cell proliferation and inhibiting cancer cell apoptosis, via targeting TP53AIP1.
35
In this case,
TP53AIP1 is determined as a helpful gene in the inhibition of tumor growth, and its upregulation
in SCC indicated that the role of NMI could be contrary to miR‑505‑5p, which is a helper, not an
38
inhibitor. The correlation implied that the increased TP53AIP and sensitized NMI facilitated each
other to decrease cell proliferation and accelerate cell apoptosis, which is worth more investigation.
3.3.8 TMC5
Transmembrane Channel Like 5, is a protein-coding gene. It was reported upregulation in
ADC and downregulation in SCC, thus TMC5 was treated as a biomarker in ADC.
27
There were
6 out of 9 NSCLC cell lines derived from ADC subtype, if supposing that the overall TMC5
expression was upregulated among all these cell lines, then TMC5 resisted NMI’s treatment
(Figure 12K), but did not show correlation with MAO A (R = -0.07). This result indicated that
NMI does not work on NSCLC related to TMC5, and conversely, its efficacy could be negatively
affected by TMC5. Interestingly, TMC5 was also significantly upregulated in prostate cancer.
Wanfeng et al. found that TMC5 knockdown significantly inhibited prostate cancer cell
proliferation via arresting cell cycle at G1 phase, thus it can be concluded that the mechanism of
NMI does not involve in the G1 phase cell cycle.
36
Therefore, a combination of TMC5 inhibitor
with NMI might achieve better therapeutic efficacy in treating NSCLC and prostate cancer.
3.3.9 RHPN2
Rhophilin Rho GTPase Binding Protein 2 is a protein-encoding gene that is located on
chromosome 19q12-13 and encodes a member of the rhophilin family of Ras-homologous (Rho)-
GTPase binding proteins, which has been shown to promote tumor progression in various types of
cancer. Yan et al. found that RHPN2 is overexpressed in NSCLC and it is a target for miR-200a.
Their study proposed that miR-200a suppressed tumor in NSCLC by targeting RHPN2.
37
Figure
12L showed that RHPN2 is resistant to NMI, this result could be proved by the gene-gene
correlation that RHPN2 is an MAO A-related gene, with R = -0.56, its overexpression decreased
MAO A’s expression, therefore weakening NMI’s targeting. Similar to TMC5, RHPN2 was
39
upregulated in prostate cancer
37, 38
, thus a combination of RHPN2 inhibitor with NMI might
strengthen the efficacy of NMI in treating NSCLC and prostate cancer.
In summary, as shown in Table 3, some genes exhibited similar correlation patterns:
CDHR2, SLC14A2, CRNN, SCGN, SOX10, TP53A1 were all upregulated in NSCLC and had
negative correlations with NMI, which means that NMI’s sensitivity elevated with the
overexpression of these genes, and they had weak or no correlation with MAO A. Based on their
mechanism, CRNN and SCGN promoted tumor cell proliferation, therefore NMI is likely to show
its efficacy by targeting these genes; TP53A1 promoted p53-dependent apoptosis, thus the
treatment of NMI may facilitate TP53A1 to decrease cell proliferation and accelerate cell apoptosis;
the mechanisms of CDHR2, SLC14A2 and SOX10 were still unknown or undetermined, so the
explanation of their high correlation with NMI could be either way.
Besides, TMC5 and RHPN2 were also similar in that their upregulation weakened NMI.
As their mechanism, they were both tumor promoters, which indicated that NMI did not treat
NSCLC via these 2 genes. On the contrary, their expression may inhibit NMI’s effectiveness,
especially for RHPN2, whose expression also inhibited MAO A, indicating that NMI should have
other mechanisms, and maybe this is the reason that MAO A-NMI correlation was not strong. A
combination of NMI with an inhibitor of these genes could probably achieve better therapeutic
efficacy. The last gene, IQCA1, expressed a unique pattern that the downregulation of IQCA1
decreased the sensitivity of NMI, whereas IQCA1 is also negatively correlated with MAO A, a
possible guess of the efficacy of NMI is that its downregulation resulted in MAO A’s upregulation,
and NMI can still work eventually as a MAO A inhibitor.
The summary of these correlations implied that both of the hypotheses may be correct. In
addition to MAO A inhibition, NMI might play a role in regulating other genes. Also, the
40
expression of some genes might influence the regulation of MAO A, which means more genes
were involved in the mechanism of NMI.
Table 3 Correlations of 9 genes with NMI, MAO A, their regulation on NSCLC, and their
mechanisms.
Gene
NSCLC
Regulation
NMI
Correlation
MAO A
Correlation
Mechanism
CDHR2 +
-0.74
High
-0.07
No
/
SLC14A2 +
-0.71
High
0.32
Weak
/
IQCA1 –
-0.68
Moderate
-0.52
Moderate
/
CRNN +
-0.65
Moderate
-0.33
Weak
Cell proliferation promoter via cyclin
D1 inducing, PI3K/AKT activator
SCGN +
-0.64
Moderate
-0.32
Weak
Cell apoptosis inhibitor, cell cycle of
G0/GI1 phase promoter
SOX10 +
-0.64
Moderate
-0.36
Weak
Tumor suppressor via Wnt/β-catenin
pathway & Tumor promoter
TP53AIP1 +
-0.63
Moderate
-0.36
Weak
Mediating p53-dependent apoptosis
TMC5
+ (ADC)
– (SCC)
0.75
High
-0.07
No
Cell cycle at G1 phase promoter
RHPN2 +
0.72
High
-0.56
Moderate
Target for miR-200a, tumor
progression promoter
Gene regulation was represented as “+” (upregulation) and “–” (downregulation), R of 0 - 0.3 (0
to -0.3) represented no correlation; R of 0.3 - 0.5 (-0.3 to -0.5) represented weak correlation; R of
0.5 - 0.7 (-0.5 to -0.7) represented moderate correlation; R of 0.7 - 0.9 (-0.7 to -0.9) represented
high correlation; and R of 0.9 – 1.0 (-0.9 to -1.0) represented very high.
NSCLC, non-small cell lung cancer; MAO A, monoamine oxidase A.

41
Discussion
As the previous results that were shown in Table 2, NMI may have a unique mechanism
to NSCLC treatment compared with 15 FDA-approved NSCLC drugs. In this case, expanding the
comparing objects, from only NSCLC drugs to all anti-cancer drugs in the DTP database should
be helpful in exploring its mechanism of action. The comparison expansion was divided into 2
sections, marketed drugs and unmarketed drugs. The comparison methodology was the same.
Among all currently marketed anti-cancer drugs contained in the DTP’s database, 10 drugs
with the highest correlation with GI50 of NMI were listed in Table 4, and their PCC of TGI and
LC50 were also listed. In terms of the DTP, an absolute value of PCC < 0.7 was regarded as no
correlation, and a PCC > 0.8 or PCC < -0.8 was regarded as a very strong correlation with the
range of PCC from -1 to 1. The result of GI50 of NMI (NSC S791228, shown on the last row of
the original heatmap) showed that the correlations were from 0.55 to 0.65, even the drug with the
highest PCC (Dactinomycin) was less than 0.7, indicating these patterns were just moderately
similar with NMI. What’s more, some drugs with relatively good PCCs of GI50 showed weak
PCCs of TGI or LC50 though. Among 10 drugs, only two drugs, Homoharringtonine and
Carfilzomib, have fine PCCs of three parameters, with PCC of GI50 = 0.58, PCC of TGI = 0.72,
PCC of LC50 = 0.53 (Homoharringtonine), and PCC of GI50 = 0.63, PCC of TGI = 0.62, PCC of
LC50 = 0.43 (Carfilzomib).
Homoharringtonine is a plant alkaloid, it was approved by FDA for the treatment of chronic
myeloid leukemia with resistance and/or intolerance to imatinib or other tyrosine kinase inhibitors.
The mechanism of Homoharringtonine was considered to be that it first fixed to the ribosomes,
and then disabled the nascent peptide chain to elongate, therefore inhibiting the synthesis of protein.
A study of Homoharringtonine’s efficacy on acute myeloid leukemia stated that the mechanism
42
was DNA epigenome-related, via the suppression of the SP1/TET1/5hmC/FLT3/MYC signaling
pathways. Another study found that Homoharringtonine also worked on triple-negative breast
cancers via probably the inhibition of anti-apoptotic proteins Mcl-1 and Bcl-2.
39, 40

Table 4 Top 10 pairwise PCCs of NMI with all marketed FDA-approved drugs by COMPARE.

PCC ranged from -1 to 1. A PCC of 1.0 represented a perfect correlation, a PCC of -1.0 represented
a perfect mirror correlation, whereas a PCC of 0 represented there was no correlation between the
two patterns. |PCC| > 0.8| was regarded as strong correlation. Five-dose concentrations for drugs
ranged all from 10
-8
to 10
-4
M.
PCC, Pearson Correlation Coefficient; NSCLC, non-small cell lung cancer; GI50, growth
inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %.

43
Carfilzomib is a modified epoxyketone that selectively inhibits the proteasome enzymes.
It was approved by FDA for or the treatment of multiple myeloma patients who have received
lines of therapy alone or in combination. Mechanically, Carfilzomib is functional because it
irreversibly and selectively binds to the proteasomes, thus delaying cell proliferation and inducing
cell apoptosis. A study reported that Carfilzomib could respond to chronic lymphocytic leukemia
by alteration of the ubiquitin proteasome pathway. It induces the accumulation of
CCAAT/enhancer-binding protein homology protein, Mcl-1, and Noxa, and relieved the Mcl-1’s
protective effect on tumor cells.
41, 42

The unmarketed drugs were listed in Table 5. The status of these drugs is all before the
approval process of FDA, some agents are currently on clinical trials, and some are just preclinical
compounds, waiting for more lab work. By selecting the GI50 of PCC greater than 0.7, a total of
16 drugs was shown. The results indicated that more unmarketed drugs could have similar patterns
to NMI, among them, 2 drugs showed overall strong correlations, NSC 800374 (PCCs for GI50 =
0.76, TGI = 0.7, LC50 = 0.63), and NSC 36359 (PCCs for GI50 = 0.71, TGI = 0.69, LC50 = 0.69),
and 5 drugs, NSC 720622, NSC 762152, NSC 269754, NSC 682066, NSC 8672, showed fine
correlations. Since these agents are still under investigation, the information about them is
extremely rare, barely any details could be found.
According to few available data, NSC 762152 is an ONX-0914 TFA salt that is under
clinical trial now. It is a third‐generation immunoproteasome‐specific inhibitor for the highly
active immunoproteasome subunit LMP7, which acted potency for the treatment of severe virus‐
mediated inflammation of the heart. A recent study speculated that ONX-0914 inhibited
phosphatase and tensin homolog on chromosome ten (PTEN) degradation, thereby resulting in the
inhibition of multiple signals including AKT/mTOR, ERK1/2, transforming growth factor-β, and
44
IKB/NF-kB. ONX-0914 blocks PTEN degradation and activates its downstream mediators, thus
weakening the Ang II-induced cardiac hypertrophy.
43, 44

Table 5 Top pairwise PCCs of NMI with all unmarketed drugs by COMPARE.

PCC ranged from -1 to 1. A PCC of 1.0 represented a perfect correlation, a PCC of -1.0 represented
a perfect mirror correlation, whereas a PCC of 0 represented there was no correlation between the
two patterns. |PCC| > 0.8 was regarded as strong correlation. Five-dose concentrations for drugs
ranged all from 10
-8
to 10
-4
M.
PCC, Pearson Correlation Coefficient; NSCLC, non-small cell lung cancer; GI50, growth
inhibition of 50 %; TGI, total growth inhibition; LC50, lethal concentration of 50 %.

45
Based on the above information, there are some surprising findings:
1) Homoharringtonine and Carfilzomib were both utilized to treat leukemia, and they were
all related to the anti-apoptotic proteins Mcl-1, which is partially consistent with NMI’s treating
efficacy that NMI performed a very potent GI50 to leukemia cell lines (Figure 5A, p < 0.01;
Figure 6A, 5 out of 6 leukemia cell lines were highly sensitive);
2) Carfilzomib and ONX-0914 were both proteasome inhibitors.
Broadly speaking, their mechanisms are crossed via relation to protein synthesis inhibition,
which could be a hint that NMI might also be involved in protein synthesis inhibition since their
patterns were so similar. What’s more, the existing intersection suggested that NMI probably can
be used to treat leukemia, or even blood disease, with a study showing that the MAO A expression
was increased in children with acute lymphoblastic leukemia.
45

Besides, NSC 269754 is a modified new trichothecene, more specifically, a Verrucarin A,
belonging to the type D trichothecene. NSC 682066 is a phosphonium porphyrin, NSC 8672 is
called Hofmann's violet, it was composed of triethylrosaniline hydrochloride and used as a
biological stain. They are such different agents with various chemical structures and treatment
targets while exhibiting high similarity with NMI, which means that the mechanism of NMI is
complicated and needs further investigation.
One more support for NMI’s treating capability is that in Figure 5, the overall comparison
between all types of cancer showed that melanoma responded well on three parameters to NMI,
and in Figure 6, most Melanoma cell lines are sensitive to the treatment of NMI. Although
melanoma is an MAO A-decreased cancer, previous results have suggested that NMI did not work
as MAO A inhibitor solely, it may also be effective for melanoma treatment through other
pathways.
46
As previously mentioned, the possible mechanism of MAO A is about inducing ROS, EMT,
and tumor hypoxia via PI3K/AKT/TWIST1 pathway, which is partially consistent with some
genes studied, indicating that the mechanism of NMI on NSCLC might be similar with on prostate
cancer. Also, there are supposed to exist other new mechanisms of NMI, not limited to MAO A
inhibition.
Fei et al. reported that the increased MAO A expression in NSCLC tissues was correlated
with the advanced stage NSCLC, for the positive rate of MAO A expression in stage III was higher
than stages I and II (Z= -2.596, P=0.029).
12
It is a pity that this thesis did not explore the association
between NMI and NSCLC stages, which is supposed to provide more insights for this study. But
the further investigation could be extended to the potency of NMI on different stages of NSCLC,
to verify if NMI is selective to the cancer stages.
Albeit no stage investigation was done, the subtype selectivity was mentioned a bit.
According to the analysis, it was shown that NMI did not express treating preference to a specific
subtype. One issue is that the analyzed samples were not sufficient. The NCI-60 screen program
chose 59 human tumor cell lines, among them 9 represented NSCLC. However, there was only
one cell line representing NSCLC subtype SCC, and two cell lines representing subtype LCC,
which were insufficient to draw a valid conclusion about subtype differences in my opinion. They
were the classic cell lines utilized in a large amount of study, but with the amazing rate of scientific
development, it will be more beneficial if DTP could enlarge their screening cell lines base. Or
perhaps this could also be the next step of the investigation in our lab.

47
Conclusion
Based on the analysis, our results showed that NMI is potent on NSCLC treatment, and its
mechanism may be involved in several pathways. The five-dose curve result showed the 6 cell
lines reached 100% growth inhibition and 3 cell lines reached 50% growth inhibition out of total
9 NSCLC cell lines at 10 μM (10
-5
M) treatment of NMI. The overall comparison between all types
of cancer and NSCLC showed that NSCLC had no significant difference with most of the other
cancers except melanoma, meaning that NSCLC is worth further investigation. Cell lines
sensitivity investigation showed that there was no distinct sensitivity pattern of NMI, and more
than half of NSCLC cell lines exhibited sensitivity against NMI, especially HOP-92 and NCI-
H522, suggesting that NSCLC has the potential to be treated by NMI. Additionally, NMI might
be able to treat different NSCLC subtypes because those sensitive cell lines representing different
3 subtypes of NSCLC. The potency of NMI on each cell line comparison showed that 4 out of 9
cell lines, HOP-62, HOP-92, NCI-H226, NCI-H522, exhibited great potency compared to prostate
cancer cell lines, especially NCI-H522 and HOP-92. Thus, NMI is potent on NSCLC, and more
potent on 4 cell lines than the rest of the 5 cell lines. In terms of subtypes, the results were
consistent with the z score activity study, that NMI is probably efficient for all subtypes. The NMI
and FDA-approval NSCLC drug potency comparisons showed that NMI outperformed most of the
marketed FDA-approved NSCLC drugs. The safety profile of NMI showed that the therapeutic
index of NMI on 9 cell lines were all fallen into the safety range, and 6 showed good TIs, indicating
that NMI is safe enough for treatment. When compared with other drugs, NMI’s safety profile was
relatively similar to others and even better than a few drugs. The pattern similarity results showed
that NMI probably has a unique mechanism to treat NSCLC. The gene correlation of NMI analysis
showed that in addition to MAO A inhibition, NMI might be effective by regulating other genes
48
on NSCLC via 1) inhibiting CRNN and SCGN; 2) facilitating TP53A1; 3) inhibiting or facilitating
CDHR2, SLC14A2, and SOX10; 4) other mechanisms unrelated to TMC5 and RHPN2; Also, the
expression of some genes might influence the regulation of MAO A, which means more genes
were involved in the mechanism of NMI.

49
References
1. Wu JB, Lin TP, Gallagher JD, et al. Monoamine oxidase A inhibitor-near-infrared dye
conjugate reduces prostate tumor growth. J Am Chem Soc. Feb 18 2015;137(6):2366-74.
doi:10.1021/ja512613j
2. Kushal S, Wang W, Vaikari VP, et al. Monoamine oxidase A (MAO A) inhibitors
decrease glioma progression. Oncotarget. Mar 22 2016;7(12):13842-53.
doi:10.18632/oncotarget.7283
3. Shih JC. Monoamine oxidase isoenzymes: genes, functions and targets for behavior and
cancer therapy. J Neural Transm (Vienna). Nov 2018;125(11):1553-1566. doi:10.1007/s00702-
018-1927-8
4. Yang X, Shi C, Tong R, et al. Near IR heptamethine cyanine dye-mediated cancer
imaging. Clin Cancer Res. May 15 2010;16(10):2833-44. doi:10.1158/1078-0432.ccr-10-0059
5. The Human Protein Atlas. RNA expression overview.
https://www.proteinatlas.org/ENSG00000189221-MAOA/tissue
6. Uhlén M, Fagerberg L, Hallström BM, et al. Proteomics. Tissue-based map of the human
proteome. Science. Jan 23 2015;347(6220):1260419. doi:10.1126/science.1260419
7. International Agency for Research on Cancer. CANCER TODAY: Data visualization
tools for exploring the global cancer burden in 2020. Vol. 2021.
http://gco.iarc.fr/today/data/factsheets/cancers/15-Lung-fact-sheet.pdf
8. The American Cancer Society. Lung Cancer. the American Cancer Society. Accessed
January 31, 2021. https://www.cancer.org/cancer/lung-cancer.html
9. Herbst RS, Morgensztern D, Boshoff C. The biology and management of non-small cell
lung cancer. Nature. Jan 24 2018;553(7689):446-454. doi:10.1038/nature25183
50
10. Kuner R, Muley T, Meister M, et al. Global gene expression analysis reveals specific
patterns of cell junctions in non-small cell lung cancer subtypes. Lung Cancer. Jan
2009;63(1):32-8. doi:10.1016/j.lungcan.2008.03.033
11. Park SJ, Yoon BH, Kim SK, Kim SY. GENT2: an updated gene expression database for
normal and tumor tissues. BMC Med Genomics. Jul 11 2019;12(Suppl 5):101.
doi:10.1186/s12920-019-0514-7
12. Liu F, Hu L, Ma Y, et al. Increased expression of monoamine oxidase A is associated
with epithelial to mesenchymal transition and clinicopathological features in non-small cell lung
cancer. Oncol Lett. Mar 2018;15(3):3245-3251. doi:10.3892/ol.2017.7683
13. Gentles AJ, Newman AM, Liu CL, et al. The prognostic landscape of genes and
infiltrating immune cells across human cancers. Nat Med. Aug 2015;21(8):938-945.
doi:10.1038/nm.3909
14. Tseng YT, Li W, Chen CH, et al. IIIDB: a database for isoform-isoform interactions and
isoform network modules. BMC Genomics. 2015;16 Suppl 2(Suppl 2):S10. doi:10.1186/1471-
2164-16-s2-s10
15. Papatheodorou I, Moreno P, Manning J, et al. Expression Atlas update: from tissues to
single cells. Nucleic Acids Res. Jan 8 2020;48(D1):D77-d83. doi:10.1093/nar/gkz947
16. Duma N, Santana-Davila R, Molina JR. Non-Small Cell Lung Cancer: Epidemiology,
Screening, Diagnosis, and Treatment. Mayo Clin Proc. Aug 2019;94(8):1623-1640.
doi:10.1016/j.mayocp.2019.01.013
17. Shoemaker RH. The NCI60 human tumour cell line anticancer drug screen. Nat Rev
Cancer. 2006:813-23. vol. 10.
51
18. Paull K, Hamel E, Malspeis L. The prediction of biochemical mechanism of action from
the in vitro antitumor screen of the National Cancer Institute. Cancer Chemotherapeutic Agents.
1995:9-45.
19. Marshall EA, Sage AP, Ng KW, et al. Small non-coding RNA transcriptome of the NCI-
60 cell line panel. Sci Data. Oct 24 2017;4:170157. doi:10.1038/sdata.2017.157
20. Patro S, Sahu KK. Normalization: A preprocessing stage. arXiv preprint
arXiv:150306462. 2015;
21. Sereti E, Tsimplouli C, Kalaitsidou E, Sakellaridis N, Dimas K. Study of the Relationship
between Sigma Receptor Expression Levels and Some Common Sigma Ligand Activity in
Cancer Using Human Cancer Cell Lines of the NCI-60 Cell Line Panel. Biomedicines. Jan 5
2021;9(1)doi:10.3390/biomedicines9010038
22. Elyes D, Joga G, Vijay I. Pharmacometric Approach To Define Narrow Therapeutic
Index (NTI) Drugs & Evaluate Bioequivalence (BE) Criteria for NTI Drugs. presented at:
kickoff meeting of the International Partnership for Nuclear Disarmament Verification; 2015;
Washington DC, USA. https://www.fda.gov/media/108490/download
23. Holbeck SL, Collins JM, Doroshow JH. Analysis of Food and Drug Administration-
approved anticancer agents in the NCI60 panel of human tumor cell lines. Mol Cancer Ther.
May 2010;9(5):1451-60. doi:10.1158/1535-7163.mct-10-0106
24. Kent WJ, Sugnet CW, Furey TS, et al. The human genome browser at UCSC. Genome
Res. Jun 2002;12(6):996-1006. doi:10.1101/gr.229102
25. Schober P, Boer C, Schwarte LA. Correlation Coefficients: Appropriate Use and
Interpretation. Anesth Analg. May 2018;126(5):1763-1768. doi:10.1213/ane.0000000000002864
52
26. Shriwash N, Singh P, Arora S, Ali SM, Ali S, Dohare R. Identification of differentially
expressed genes in small and non-small cell lung cancer based on meta-analysis of mRNA.
Heliyon. Jun 2019;5(6):e01707. doi:10.1016/j.heliyon.2019.e01707
27. Venugopal N, Yeh J, Kodeboyina SK, et al. Differences in the early stage gene
expression profiles of lung adenocarcinoma and lung squamous cell carcinoma. Oncol Lett. Dec
2019;18(6):6572-6582. doi:10.3892/ol.2019.11013
28. Qin Y, Conley AP, Grimm EA, Roszik J. A tool for discovering drug sensitivity and gene
expression associations in cancer cells. PLoS One. 2017;12(4):e0176763.
doi:10.1371/journal.pone.0176763
29. Shayakul C, Clémençon B, Hediger MA. The urea transporter family (SLC14):
physiological, pathological and structural aspects. Mol Aspects Med. Apr-Jun 2013;34(2-3):313-
22. doi:10.1016/j.mam.2012.12.003
30. Ferrucci V, Asadzadeh F, Collina F, et al. Prune-1 drives polarization of tumor-associated
macrophages (TAMs) within the lung metastatic niche in triple-negative breast cancer. iScience.
Jan 22 2021;24(1):101938. doi:10.1016/j.isci.2020.101938
31. Li C, Xiao L, Jia J, et al. Cornulin Is Induced in Psoriasis Lesions and Promotes
Keratinocyte Proliferation via Phosphoinositide 3-Kinase/Akt Pathways. J Invest Dermatol. Jan
2019;139(1):71-80. doi:10.1016/j.jid.2018.06.184
32. Bai Y, Sun Y, Peng J, et al. Overexpression of secretagogin inhibits cell apoptosis and
induces chemoresistance in small cell lung cancer under the regulation of miR-494. Oncotarget.
Sep 15 2014;5(17):7760-75. doi:10.18632/oncotarget.2305
53
33. Tong X, Li L, Li X, et al. SOX10, a novel HMG-box-containing tumor suppressor,
inhibits growth and metastasis of digestive cancers by suppressing the Wnt/β-catenin pathway.
Oncotarget. Nov 15 2014;5(21):10571-83. doi:10.18632/oncotarget.2512
34. Cronin JC, Loftus SK, Baxter LL, Swatkoski S, Gucek M, Pavan WJ. Identification and
functional analysis of SOX10 phosphorylation sites in melanoma. PLoS One.
2018;13(1):e0190834. doi:10.1371/journal.pone.0190834
35. Fang H, Liu Y, He Y, et al. Extracellular vesicle‑delivered miR‑505‑5p, as a diagnostic
biomarker of early lung adenocarcinoma, inhibits cell apoptosis by targeting TP53AIP1. Int J
Oncol. May 2019;54(5):1821-1832. doi:10.3892/ijo.2019.4738
36. Zhang W, Wang S, Zhang X, et al. Transmembrane Channel-Like 5 (TMC5) promotes
prostate cancer cell proliferation through cell cycle regulation. Biochimie. Oct 2019;165:115-
122. doi:10.1016/j.biochi.2019.07.017
37. Huang Y, Bao T, Li Z, Ji G, Zhang L. Function of miR-200a in proliferation and
apoptosis of non-small cell lung cancer cells. Oncol Lett. Aug 2020;20(2):1256-1262.
doi:10.3892/ol.2020.11649
38. Jiang S, Mo C, Guo S, Zhuang J, Huang B, Mao X. Human bone marrow mesenchymal
stem cells-derived microRNA-205-containing exosomes impede the progression of prostate
cancer through suppression of RHPN2. J Exp Clin Cancer Res. Dec 17 2019;38(1):495.
doi:10.1186/s13046-019-1488-1
39. Yakhni M, Briat A, El Guerrab A, et al. Homoharringtonine, an approved anti-leukemia
drug, suppresses triple negative breast cancer growth through a rapid reduction of anti-apoptotic
protein abundance. Am J Cancer Res. 2019;9(5):1043-1060.
54
40. Li C, Dong L, Su R, et al. Homoharringtonine exhibits potent anti-tumor effect and
modulates DNA epigenome in acute myeloid leukemia by targeting SP1/TET1/5hmC.
Haematologica. Jan 2020;105(1):148-160. doi:10.3324/haematol.2018.208835
41. Perel G, Bliss J, Thomas CM. Carfilzomib (Kyprolis): A Novel Proteasome Inhibitor for
Relapsed And/or Refractory Multiple Myeloma. P t. May 2016;41(5):303-7.
42. Lamothe B, Wierda WG, Keating MJ, Gandhi V. Carfilzomib Triggers Cell Death in
Chronic Lymphocytic Leukemia by Inducing Proapoptotic and Endoplasmic Reticulum Stress
Responses. Clin Cancer Res. Sep 15 2016;22(18):4712-26. doi:10.1158/1078-0432.ccr-15-2522
43. Althof N, Goetzke CC, Kespohl M, et al. The immunoproteasome-specific inhibitor ONX
0914 reverses susceptibility to acute viral myocarditis. EMBO Mol Med. Feb 2018;10(2):200-
218. doi:10.15252/emmm.201708089
44. Xie X, Wang HX, Li N, et al. Selective Inhibition of the Immunoproteasome β5i Prevents
PTEN Degradation and Attenuates Cardiac Hypertrophy. Front Pharmacol. 2020;11:885.
doi:10.3389/fphar.2020.00885
45. Al-Helaly LA, Saleh SS, Bdaiwi LF. STUDYING THE ENZYMES LEVELS IN
CEREBROSPINAL FLUID FOR CHILDREN WITH ACUTE LYMPHOBLASTIC
LEUKEMIA. 2017;

Abstract (if available)

Abstract According to our previous work, prostate cancer and glioma showed increased expression levels of monoamine oxidase A (MAO A), and near-infrared (NIR) dye conjugate MAO A inhibitor (NMI) is effective for the treatment of both cancers. This study investigated the potency of NMI on another MAO A-increased cancer, non-small cell lung cancer (NSCLC), and its potential mechanism via bioinformatics analysis. ❧ Based on the NCI-60 screen data, results showed that 6 cell lines reached 100% growth inhibition and 3 cell lines reached 50% growth inhibition out of 9 NSCLC cell lines at 10 μM NMI. Among these 9 cell lines, more than half exhibited sensitivity against NMI, the sensitivity was determined by the Z score of GI50, TGI, and LC50 values. Additionally, NMI showed especially great potency on 4 out of 9 NSCLC cell lines. Thus NMI has the great potential for NSCLC treatment. Next, comparisons of the potency between NMI and other FDA-approval NSCLC drugs showed that NMI outperformed most of the marketed drugs. What’s more, the safety profile showed that NMI is safe because its therapeutic index on all 9 cell lines were in the safety range. Finally, the COMPARE algorithm of mechanism similarity showed that NMI may have a unique mechanism compared with other FDA-approved drugs. The gene correlation of NMI analysis showed that in addition to MAO A, NMI might regulate several other genes. The interaction of these genes and MAO A may be the mechanism of NMI function. In summary, this study showed that NMI is a potent drug for NSCLC, whose mechanism is unique compared to existing drugs. In addition to MAO A inhibition, there may be more genes involved in the mechanism of NMI, which warrants further investigation.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

NMI (near-infrared dye conjugate MAO A inhibitor) outperformed FDA-approved prostate cancer drugs with a unique mechanism based on bioinformatic analysis of NCI60 screening data

PDF

Study of a novel near-infrared conjugated MAOA inhibitor, NMI, against CNS cancer by NCI60 data analysis

PDF

NMI: a near infrared conjugated MAO-A inhibitor as a novel targeted therapy for colorectal and other cancers

PDF

Role of inflammation in prostate carcinogenesis and prostate cancer growth

PDF

Assessment of theranostic agent for brain cancer

PDF

Monoamine oxidase and cancer

PDF

Monoamine oxidase A inhibitors and androgen receptor antagonists regulate mitochondrial function in prostate cancer cells

PDF

Mechanism of action of rapamycin and its applications in aging, cancer therapy and metabolism

PDF

Co-expression of monoamine oxidase A and prostate cancer stem cell markers in Pten knockout mice

PDF

Potential therapeutic effect of monoamine oxidase (MAO) inhibitor on human neuroblastoma

PDF

Inhibition of MAO-A by Dual MAO-A/HDAC inhibitors: in silico approach for ligand binding and affinity prediction

PDF

Characterization of the transferrin oligomer and its potential application in drug delivery

PDF

Deregulation of CD36 expression in cancer presents a potential targeting therapeutic opportunity

PDF

Genome mining of secondary metabolites in Aspergillus terreus isolated from the Chernobyl Collection

PDF

MAO a deficient mice exhibit an altered immune system in the brain and prostate

PDF

Characterization of actin based motility in mammalian cells through LIM and SH3 domain protein 1 (LASP1) and elastin like polypeptide (ELP) fusion protein

PDF

Explore the influence of CYP21A2 silencing on H295R human adrenal cortical carcinoma cell gene expression function

PDF

Targeting cellular redox modulations for pancreatic cancer treatment

PDF

Methodology of exploring the role of SOX9 in the human HCC stem-like cells

PDF

Methods and protocols for detecting the intracellular assembly of elastin-like polypeptides

Asset Metadata

Creator Lian, Yuxuan (author)

Core Title Bioinformatics analysis of the anti-cancer potency of NMI on non-small cell lung cancer and its potential mechanism

School School of Pharmacy

Degree Master of Science

Degree Program Pharmaceutical Sciences

Publication Date 04/24/2021

Defense Date 03/18/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag MAO A,mechanism,NCI-60 screening,NMI,NSCLC,OAI-PMH Harvest

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Shih, Jean C. (committee chair), Haworth, Ian S. (committee member), Wang, Clay C.C. (committee member)

Creator Email evelian599@gmail.com,lianyuxu@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-452281

Unique identifier UC11668437

Identifier etd-LianYuxuan-9524.pdf (filename),usctheses-c89-452281 (legacy record id)

Legacy Identifier etd-LianYuxuan-9524.pdf

Dmrecord 452281

Document Type Thesis

Rights Lian, Yuxuan

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA