(NON SOLL’S
ELSEVIER
01010
0010101010010
COMPUTATIONAL
001-01010010201010101011
10
101001
0101016
41
01
0101001
1010101
10
ANDSTRUCTURAL
11 0101001
1010101 10
10 0101001 0101010
11
BIOTECHNOLOGY
00
10100
5101010
11
010101010011010107
10
110101010010010000010010
JOURNAL
journal homepage: www.elsevier.com/locate/csbj
Integrative omics analysis reveals relationships of genes with synthetic lethal interactions through a pan-cancer analysis
Check for updates
Li Guo ª, Sunjing Li ª, Bowen Qian ª, Youquan Wanga, Rui Duan b, Wenwen Jianga, Yihao Kanga, Yuyang Dou ª, Guowei Yang ª, Lulu Shen b, Jun Wanga, Tingming Liang b,c,*
a Department of Bioinformatics, Smart Health Big Data Analysis and Location Services Engineering Lab of Jiangsu Province, School of Geographic and Biologic Information, Nanjing University of Posts and Telecommunications, Nanjing 210023, China
b Jiangsu Key Laboratory for Molecular and Medical Biotechnology, School of Life Science, Nanjing Normal University, Nanjing 210023, China
“Changzhou Institute of Innovation and Development, Nanjing Normal University, Nanjing 210023, China
ARTICLE INFO
Article history: Received 9 May 2020 Received in revised form 10 October 2020 Accepted 12 October 2020 Available online 21 October 2020
Keywords: Synthetic lethality Cancer therapy Pan-cancer analysis RNA interaction
ABSTRACT
Synthetic lethality is thought to play an important role in anticancer therapies. Herein, to understand the potential distributions and relationships between synthetic lethal interactions between genes, especially for pairs deriving from different sources, we performed an integrative analysis of genes at multiple molecular levels. Based on inter-species phylogenetic conservation of synthetic lethal interactions, gene pairs from yeast and humans were analyzed; a total of 37,588 candidate gene pairs containing 7,816 genes were collected. Of these, 49.74% of genes had 2-10 interactions, 22.93% were involved in hallmarks of cancer, and 21.61% were identified as core essential genes. Many genes were shown to have important biological roles via functional enrichment analysis, and 65 were identified as potentially crucial in the pathophysiology of cancer. Gene pairs with dysregulated expression patterns had higher prognostic val- ues. Further screening based on mutation and expression levels showed that remaining gene pairs were mainly derived from human predicted or validated pairs, while most predicted pairs from yeast were fil- tered from analysis. Genes with synthetic lethality were further analyzed with their interactive microRNAs (miRNAs) at the isomiR level which have been widely studied as negatively regulatory mole- cules. The miRNA-mRNA interaction network revealed that many synthetic lethal genes contributed to the cell cycle (seven of 12 genes), cancer pathways (five of 12 genes), oocyte meiosis, the p53 signaling pathway, and hallmarks of cancer. Our study contributes to the understanding of synthetic lethal inter- actions and promotes the application of genetic interactions in further cancer precision medicine.
@ 2020 The Author(s). Published by Elsevier B.V. on behalf of Research Network of Computational and Structural Biotechnology. This is an open access article under the CC BY license (http://creativecommons. org/licenses/by/4.0/).
Abbreviations: ACC, adrenocortical carcinoma; BLCA, bladder urothelial carci- noma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL, cholangiocarcinoma; COAD, colon ade- nocarcinoma; DLBC, lymphoid neoplasm diffuse large B-cell lymphoma; ESCA, esophageal carcinoma; GBM, glioblastoma multiforme; HNSC, head and neck squamous cell carcinoma; KICH, kidney chromophobe; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LAML, acute myeloid leukemia; LIHC, liver hepatocellular carcinoma; LGG, brain lower grade glioma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; MESO, me- sothelioma; OV, ovarian serous cystadenocarcinoma; PAAD, pancreatic adenocar- cinoma; PCPG, pheochromocytoma and paraganglioma; PRAD, prostate adenocarcinoma; READ, rectum adenocarcinoma; SARC, sarcoma; SKCM, skin cutaneous melanoma; STAD, stomach adenocarcinoma; TGCT, testicular germ cell tumors; THCA, thyroid carcinoma; THYM, thymoma; TSG, tumor suppressor gene; UCEC, uterine corpus endometrial carcinoma; UCS, uterine carcinosarcoma; UVM, uveal melanoma.
* Corresponding author at: School of Life Science, Nanjing Normal University, Nanjing 210023, China.
E-mail address: tmliang@njnu.edu.cn (T. Liang).
1. Introduction
Cancer is one of the leading causes of death worldwide but many patients with metastatic cancers cannot be treated because of drug resistance [1,2]. Recently, however, a type of genetic interaction known as synthetic lethality that was first identified in studies in fruit flies [3,4] and yeast [5,6] has emerged as a promising anticancer strategy. A synthetic lethal interaction between two paired genes indicates that perturbation of either gene alone is viable, but that perturbation of both genes simulta- neously causes the loss of viability [7] (Fig. 1A). The negative genetic interaction, synthetic lethal interaction, or sick genetic interaction may be used to identify new antibiotic or therapeutic targets [8,9], and has become a potential strategy for clinical anti- cancer therapies.
https://doi.org/10.1016/j.csbj.2020.10.015
A
B
Homologous
Conservation
&
SynLethDB database
A
B
Yeast
Human
Wildtype, viable
Synthetic lethal interactions
Synthetic lethal interactions
Essential
non-essential
Essential and
Essential
non-essential
Essential and
A
x
A
100
597
282
368
661
1,671
B
B
x
0
0
1,091
0
0
1,292
Single mutation, viable
293
2,528
Non-essential
Gene number
Non-essential
Gene number
1.01
150
A
x
₡ 0.8-
B
-log10p
Interaction score
100
x
0.6-
50
0.4
0.2
Double mutation, lethal
0.
-1.1
1 -0.9 -0.7 -0.5
-0.8
-0.6
-0.4
-1.1
Interaction score essential
Interaction score essential and non-essential
5.9 -0.7 -0
-0.5
0.0
Interaction score non-essential
Gene pairs essential
Gene pairs essential and non-essential
Gene pairs non-essential
A
B
Drug
Mutation
Expression
NcRNA
*_
Kill tumor cell via targeting interacted gene
Gene pairs with synthetic lethal interactions
In several human cancers, novel therapeutic strategies are rapidly developing based on interactions of synthetic lethality via the exploitation of loss-of-function mutations [10]. Mutant combi- nations can be queried to screen and identify potential synthetic lethal interactions, but limited synthetic lethal interactions with higher confidence levels may hinder the possibility of developing therapeutic targets. Compared with humans, largescale screening of model organisms enables the straightforward surveillance of multiple potential synthetic lethal interactions. This has been sys- tematically studied and validated in yeast, and high conservations of genetic interactions [11-16] have enabled the identification of candidate gene pairs via phylogenetic conservation. Predictions of cross-species genetic interactions may provide more references for identifying potential cancer-relevant synthetic lethal interac- tions, which would allow the specific targeting of cancer cells. Although prediction by validated synthetic lethal interactions in model organisms may provide more data references for cancer treatment, it is nevertheless important to understand the potential features of these predicted gene pairs, especially those identified via integrative analysis.
In this study, to determine potential correlations between pre- dicted gene pairs from yeast and humans, we performed a system- atic pan-cancer analysis at multiple molecular levels based on collected synthetic lethal interactions. These mainly included pre- dicted gene pairs from yeast based on evolutionary conservation and predicted or verified gene pairs from humans. The potential relationships of candidate gene pairs were surveyed at the muta- tion and expression levels across a diverse range of cancer types.
Additionally, in-depth analyses of screened gene pairs were per- formed, including the identification of potential therapeutic values for further cancer treatment and potential interactions with nega- tive regulatory microRNAs (miRNAs). Several studies have shown the existence of multiple isomiRs in miRNA [17-20], which are heterogenous with respect to sequence, length, and expression. We therefore mainly investigated miRNA-mRNA interactions at the isomiR level. Our integrated analysis provides an understand- ing of the relationships of paired genes with synthetic lethal inter- actions, which will facilitate the identification of mechanistic complexities with potential applications in anticancer therapies.
2. Materials and methods
2.1. Data resources
Candidate synthetic lethality interactions were first collected according to predicted gene pairs from experimentally validated pairs in yeast [21] using InParanoid 6 [22] based on evolutionary conservation (http://inparanoid.sbc.su.se/cgi-bin/index.cgi) (Fig. 1B). Genes were collected based on their phylogenetic conser- vation, and were always ancient genes in the evolutionary process. Because novel genes are also important in cancer pathophysiolog ical processes [23], we simultaneously collected human candidate predicted or validated synthetic lethality interactions from the SynLethDB database [24] (Figs. 1 and S1).
To perform multiple analyses of these collected candidate gene pairs, we obtained mutation data, gene expression profiles, small
RNA expression profiles, and relevant clinical data for a diverse range of cancer types from The Cancer Genome Atlas (TCGA) (https://tcga-data.nci.nih.gov/tcga/) using the “TCGAbiolinks” package [25]. Involved gene pairs were queried for detailed drug responses using the Genomics of Drug Sensitivity in Cancer data- base (GDSC) [26] (|DF| > 0.10 and p < 0.05 were considered signif- icant correlations).
2.2. Functional enrichment analysis and potential gene characteristics in tumorigenesis
To understand potential biological functions of candidate gene pairs, relevant genes were analyzed using The Database for Anno- tation, Visualization and Integrated Discovery (DAVID) version 6.8 [27]. Further, z scores in DAVID were estimated using the following formula based on expression patterns in breast invasive carcinoma (BRCA), which was used as an example to understand expression trends:
z score = (up - down) ☒ Vcount
where up and down are the numbers of up-regulated and down- regulated genes in BRCA, respectively, and count indicates the total gene number.
These genes were also queried for their potential roles in cancer physiology, based on the distribution of hallmarks of cancer [28] (http://software.broadinstitute.org/gsea/msigdb/), genes in the cancer gene census (CGC) [29] (http://cancer.sanger.ac.uk/census), core essential genes (common genes from Hart et al. [30], Blomen et al. [12], and Wang et al. [31]), oncogenes, tumor suppressor genes [32], and actionable genes [33].
2.3. Survival analysis
To estimate the potential prognostic values of candidate gene pairs, survival analysis was performed based on two groups (both mutations (MM) and both wildtypes (WW) at the mutation level, both abnormally expressed (AA) and both normally expressed (NN) at the expression level) and three groups (MM, MW, WW; AA, AN, NN) at mutation and expression levels, respectively. A log-rank test was used to estimate the potential difference, and p < 0.05 was considered statistically significant.
2.4. Screening related regulatory miRNAs for candidate genes
Most human genes are negatively regulated by miRNAs, which play an important role in pathological processes and the occur- rence and development of cancers [34,35]. Therefore, for candidate gene pairs with synthetic lethal interactions, we further surveyed related regulatory miRNAs for each relevant gene to understand the interactions between different RNAs. First, based on screened genes, related miRNAs were mainly obtained from starBase v2.0 [36], and these miRNA-mRNA pairs were considered potential can- didate interactions between mRNAs and small non-coding RNAs (ncRNAs). Then, miRNAs with adverse expression patterns were further screened. The expression profiles of miRNAs were mainly collected from the most dominantly expressed isomiR for each miRNA locus to estimate the expression pattern of classical miR- NAs based on that of multiple isomiRs.
2.5. Randomization test
To determine the significance of detected frequencies of prog- nostic values of candidate gene pairs, a randomization test was performed by randomly selecting other gene pairs (generated by
CFinder [37]) with equal numbers. This analysis was repeated 1000 times (the significance was estimated based on the propor- tion of times) to assess whether the observed average values were higher than the actual average values.
2.6. Statistical analysis and network visualization
Abnormal expression profiles for mRNAs and miRNAs were assessed using DESeq2 [38], and hypothesis testing in relevant analysis was used to estimate the potential difference between or among groups (such as a trend test). Potential interactions between multiple genes were presented using Cytoscape 3.7.1 [39]. Venn distributions were analyzed using a publicly available tool (http://bioinformatics.psb.ugent.be/webtools/Venn/), and all statistical analyses were analyzed using R programming language (version 3.6.1).
3. Results
3.1. Overview of collected gene pairs with synthetic lethality
According to validated gene pairs with synthetic lethality in yeast (score ≤-0.35), we collected relevant genes to screen homol- ogous human gene pairs using InParanoid 6 (Fig. 1B and S1A). Involved gene were classified as essential or non-essential genes. Pairs containing essential genes were common, although their partners might not be essential genes (Fig. S1B). Additionally, the detailed gene features might not be consistent with those in yeast. Most gene pairs were scored between -0.35 and -0.80, and these were considered candidate pairs to perform further analysis.
Simultaneously, to understand the potential correlations of the predicted conserved gene pairs with humans, we also collected human gene pairs with synthetic lethal interactions from the Syn- LethDB database. Thus, a total of 37,588 candidate gene pairs con- taining 7,816 genes were obtained (Tables S1 and S2). Of these, only 1066 genes were found to be common between data from yeast and the SynLethDB database (the top picture in Fig. S1C). Compared with the specific genes collected from human gene pairs (n = 5453), fewer genes (n = 1297) were collected from yeast. Most of these genes showed abnormal expression patterns in cancers (middle picture in Fig. S1C and D), implicating their potential roles in tumorigenesis.
3.2. In-depth gene analysis showing potentially important biological roles
Most genes involved in potential synthetic lethal interactions were found to have 1-10 interactions (Fig. 2A and lower picture in Fig. S1C). Specifically, 49.74% of genes were found with 2-10 interactions, and only 2.28% of genes had more than 51 interac- tions (Fig. 2A). These direct or indirect interactions would likely complicate synthetic lethal interactions and further gene-drug interactions.
Genes with potential synthetic lethal interactions could be drug targets for cancer treatment. To understand their biological roles, we investigated their specific characteristics We found that 22.93% of these genes were involved in hallmarks of cancer, and 21.61% were identified as core essential genes (Fig. 2B and Table S2). Many genes were shown to have multiple characteristics (Fig. 2B). For example, both ABL1 and BCL2 genes were validated as oncogenes, actionable genes, essential genes, genes in CGC, poten- tial drug targets, and also contributed to hallmarks of cancer. This provided evidence for their possible roles in cancer treatment, so they were analyzed further.
A
B
hallmark
2.28%
26.17%
1,250
22.93
>51
1,155
1000
5.26
11-50
21.61
CGC
1
Frequency
Intersection Size
20
essential
500
6.53
16.03%
actionable
TSG
29.72%
18.27%
214
234
oncogeneDrug target
96
146
75
9.65%
oncogene
0
10
TSG
49.74%
actionable
CGC
0
Drug target
2-10
2
10
20
30
40
50
essential
hallmark
Distribution of number of interaction
Interacted number (2-50)
1500 500 Set Size
C
D
FUIPI
RAC1
SFPQ
ARAF
30
16
0
NF2
Number of degree
NF1
CDKNZA
NRAS
CALR
Frequency of degree
EZH2
FBXW7
HRAS
COC73
FEN1
KRAB
COH1
ATR
BRCA2
AR
20
+
PTEN
JAK1
MLH1
BRAF
TP53
MAP2K1
FLT3
12
APC
VHL
MYC
POLE
ATRX
RAD51
KSW
MTOR
BRCA1
COKE
RB1
KIT
ATM
10
BCL2
COK4
MET
AKT1
SMAD4
SMO
8
MDM2
RET
PDGFRA
Frequency
COND1
ERBB2
EĞER
RARA
AURKA
MSHG
ALK
ABL1
0
RHOA
SMARCAA
PIK3R1
CSFIR
AKT2
CDH1
COKNZA
CSFIK
FUBP
PDGFRA
SPERO
NOTCH
RHOA
SMAD2
SMADS
SMO
ABL1
AURKA
THET
EZEL
APC
CALR
CDE
MLH1
NE
RAC1
SFPO
FBXW7
MAP2K
MET
RARA
SMARGT
BRAF ATR
VHL
PTEN
KIT
BRCA2
MDM2
COKA
POLE
BCL2
ATM
RALDI
BRCA1
NRAŠ
COKE
HRAS
MTOR
TPOA
KRAS
EGFR
DDR2
FGFR1
SMAD2
FOFR2
AKT2
NOTCH1
Gene
E
BP
CC
MF
z-score increasing
10
-log FDR
decreasing
U
0
DNA repair
double-strand break repair via homologous recombination
liver regeneration
DNA synthesis involved in DNA repair
positive regulation of DNA replication
response to drug
Ras protein signal transduction
regulation of signal transduction by p53 class mediator
replicative senescence
negative regulation of cell-matrix adhesion
negative regulation of transcription, DNA-templated
positive regulation of GTPase activity
positive regulation of gene expression
positive regulation of transcription, DNA-templated
positive regulation of stress fiber assembly
ERBB2 signaling pathway
positive regulation of protein phosphorylation
cell cycle arrest
positive regulation of apoptotic process
positive regulation of MAP kinase activity
epidermal growth factor receptor signaling pathway
regulation of protein stability
response to estradiol
signal transduction
transmembrane receptor protein tyrosine kinase signaling pathway
positive regulation of cell cycle
positive regulation of epithelial cell proliferation
protein stabilization
regulation of cell motility
cellular response to UV
visual learning
negative regulation of transcription from RNA polymerase Il promoter
MAPK cascade
cell proliferation
cellular response to DNA damage stimulus
protein phosphorylation
negative regulation of G1/S transition of mitotic cell cycle
peptidyl-tyrosine phosphorylation
intrinsic apoptotic signaling pathway in response to DNA damage
positive regulation of fibroblast proliferation
phosphatidylinositol phosphorylation
negative regulation of cell proliferation
positive regulation of ERK1 and ERK2 cascade
positive regulation of cell migration
negative regulation of apoptotic process
positive regulation of transcription from RNA polymerase Il promoter
ureteric bud development
regulation of phosphatidylinositol 3-kinase signaling
protein autophosphorylation
multicellular organism growth
positive regulation of MAPK cascade
phosphatidylinositol-mediated signaling
positive regulation of cell proliferation
negative regulation of epithelial cell proliferation
in utero embryonic development
membrane
PML body
nuclear chromosome, telomeric region
nuclear chromatin
cytoplasm
nucleoplasm
nucleus
protein complex
plasma membrane
cytosol
protein kinase binding
double-stranded DNA binding
protein kinase activity
enzyme binding
chromatin binding
protein phosphatase binding
ubiquitin protein ligase binding
kinase activity
protein serine/threonine kinase activity
protein tyrosine kinase activity
protein binding
transcription factor binding
Ras guanyl-nucleotide exchange factor activity
transmembrane receptor protein tyrosine kinase activity
ATP binding
identical protein binding
phosphatidylinositol-4,5-bisphosphate 3-kinase activity
Significant GO terms
Gene interactions were shown to be quite complex based on an analysis of 65 genes that had been validated with at least four types of characteristics (Fig. 2C). Some genes were found to only interact with one other gene (n = 17, 26.15%), but most had multi- ple interactions that were quite complex (Fig. 2C and D). We only present some of the interactions from the 65 screened genes, but more widespread interactions exist within all collected genes (Fig. S2A and B). Most relevant gene pairs (each containing one or two screened genes) had three interactions (Fig. S2A), but some genes including KRAS, HRAS, and NRAS had more than 1500 interac- tions (Fig. S2B), implying their important role as hub genes. Indeed, these three genes are known to have crucial biological roles in the occurrence and development of cancers. Oncogenic KRAS drives an immune suppressive program in colorectal cancer by repressing interferon regulatory factor 2 expression [40], and may sensitize lung adenocarcinoma to GSK-J4-induced metabolic and oxidative stress [41]; moreover, KRAS-targeted anticancer strategies have been documented [42]. Additionally, HRAS-driven cancer cells are vulnerable to TRPML1 inhibition [43].
These 65 screened genes were also analyzed for their potential biological roles to help understand their function in multiple bio- logical pathways. We detected a series of significantly enriched gene ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (false discovery rate [FDR] < 0.05) (Fig. 2E and Fig. S2C), implying that most have crucial roles in mul- tiple biological processes. More importantly, a pan-cancer analysis showed that many of these genes were relatively stably expressed across a range of cancer types (Fig. S2D).
3.3. Analysis of candidate gene pairs at the mutation level
Although candidate synthetic lethal interactions were initially identified from yeast and human predicted/validated pairs, further screening was essential to obtain gene pairs with higher confi- dence levels based on an integrative analysis of multiple mole- cules. First, the mutation profiles of all involved genes was investigated in 33 cancer types. We collected a total of 75 gene pairs (containing 74 genes), and the mutation status of both the two-paired genes was detected (each gene pair was detected in at least five cancer types) (Fig. 3A). Some gene pairs had higher mutation frequencies, especially in the uterine corpus endometrial carcinoma. Missense mutations were the most common mutation type (Fig. 3A). To understand their potential value as drug targets, the 75 gene pairs were investigated for their correlations with drug response. Interestingly, some genes showed significant positive and negative correlations with the drug response in specific cancer types based on a comparison of both mutations (MM) and both wild types (WW) of the two-paired genes (Fig. 3B), MM and MW, and MW and WW gene pairs (Fig. S3A-C). Compared with compar- isons in multiple groups, more significant correlations could be found between groups of MM and WW (Fig. S3C). These results implied the potential role of the complex genetic interactions in relevant anticancer drug design.
To better understand the biological function of the these genes, functional enrichment analysis was performed using DAVID. Mul- tiple significant biological pathways were enriched, including pathways in cancer, glioma, central carbon metabolism in cancer, miRNAs in cancer, melanoma, non-small cell lung cancer, and pros- tate cancer (Fig. 3C). Many of the genes showed abnormal expres- sion patterns in some cancer types, and most showed consistent dysregulated trends across a diverse range of cancers (Fig. S3D). Interestingly, only 11 genes were predicted to be conserved in yeast, six were also found in the SynLethDB database, and 63 were obtained from human predicted or validated gene pairs (Fig. S3E). Among the six common genes, most showed relatively stable expression in a diverse range of tissues, and no significant
differences could be detected among cancer samples (Fig. S3D and E). Further analysis based on potential gene functions showed that many of them had roles in hallmarks of cancer, and some were potentially crucial in the occurrence and development of cancer (Fig. S3E).
To estimate the potential value of these synthetic lethal interac- tions, the role of gene pairs as prognostic markers was investigated based on survival analysis. Comparisons between the two groups and among the three groups were analyzed, and the gene pairs were shown to be significantly more likely to be potential prognos- tic markers than other pairs without synthetic lethal interactions based on a randomization testing (1,000 times, p = 0.035 < 0.05 for the two groups, and p = 0.040 < 0.05 for the three groups) (Fig. 3D). These results suggest that the synthetic lethal interac- tions could be markers for disease prognosis, and also indicate their importance in the development of cancer and potential roles in further drug treatment.
3.4. Analysis of candidate gene pairs at the mRNA level
Based on candidate synthetic lethal interactions, the potential expression patterns for the two-paired genes could be used as markers to estimate their expression and further biological func- tion. Therefore, we screened abnormally expressed genes from candidate gene pairs, and collected those that were dysregulated in more than 10 cancer types (Fig. 4A). Many of these genes showed consistent expression in a diverse range of cancer types, suggesting the similarity of their roles in tumorigenesis.
Compared with gene pair analysis at the mutation level, gene pairs at the mRNA level also showed more significant prognostic values than other gene combinations without potential synthetic lethality based on a randomization testing (1000 times, p < 0.001 < 0.05 for the two groups, and p = 0.012 < 0.05 for the three groups) (Fig. 4B). Interestingly, we found that paired genes both showing dysregulated expression were associated with a higher probability of long-term survival than other pairs with one gene dysregulated or both normally expressed (Fig. 4B). Similar to analysis at the mutation level, these results indicated that the synthetic lethal interactions have potential prognostic value in cancer treatment.
We also screened 97 gene pairs containing 68 dysregulated genes (paired genes were identified as dysregulated expression in more than 10 cancer types) (Fig. 4C). The interaction network showed potential interactions between these genes, with up- regulated expression patterns dominating (Fig. 4C and D). Based on whole candidate gene pairs with synthetic lethal interactions, many of these genes were found to have more complex interac- tions than expected (Fig. 4E), implicating their potential roles and interactions with drug sensitivities.
3.5. Candidate gene pairs based on mutation and expression levels
A total of 4023 candidate gene pairs were collected that included one gene with more than 2.0% mutation frequencies in at least five cancer types. The expression patterns of these gene pairs were then investigated, and 377 pairs containing 310 genes were identified in which one gene showed abnormal expression in more than 10 cancer types (Fig. 5A). Of these, only 28 were iden- tified as predicted gene pairs from yeast, and most were derived from human synthetic lethal interactions.
A total of 91 gene pairs (Table S3) were identified containing one mutated gene in at least five cancer types and its partner with up-regulated expression in more than 10 cancer types. Of these pairs, 53 were mutated in the first gene (the relative position in paired genes) and 38 were mutated in the second gene. Compared with the mutated genes, their partners showed obvious up- regulation across a diverse range of cancer types (74.90% and
A
Frequency
MF (%)
0
Frequency
0
60
8
40-
5
20-
R
OS
2
9
2
31
3
6 6
45
16
9
42
23
11
6.
16
20
20
12
25
ARHGEF2:TP53
5
6
HIPK2-TP53
L
BF
C
A
14
A
Y
2
37
Y
23
7
5 2
4
K
KRAS:POLA1
16
4
C
30
78
ATM:NLRP2
ACC
EG
P
[21
12
3
1
L
14
9
28
20
24
25
16
22
32
PRDM9:PRKCG PRKCO PROMO
3
22
23
BLCA
16
g
8
4
14
29
PRI
FLI-PLEC
BRCA
P
4
29
13
AH
14
22
18
10
FRMPD1:KRAS
CESC
48
4
11
26
4
12
DI ICER
b
12
6
28 34
9
9
DIDO1:KRAS
4
PML-TP53
CHOL
18
L
KN2
P
13
16
11 13
11
PRKDC REICCA
COAD
SKALSE
D
2
10
3
S
12
19
20
15
GRM8 KRAS
MITOR
A
5
25
12
7
DLBC
22
16
15
12
18
A
PDGEDE DE
ESCA
1
ARA
1
18
91
22
E 13
a
A
8
30
28
42
10
PCDHABE
GBM
19
23
5
18
2
11
35
14
6
10
13
ADAM
GLI2:PIKSCA
SO KRAS
HNSC
ME
N
19
12
2
15
Z
18
15
11
5
OSPANNET
PRKCB:SCN3A
KICH
8
7
28
ALMS1 KRAS
KIRC
KAL
P
RA
23
14
26
ARID
8
-
5
17
12
5
10
15
19
9 16
13
22
41
MET:TP53
KRAS:SCN4A KALRN:KRAS
KIRP
PIK
P
14
6
5
14
PIK3
8
5
13 23 28
19
LAML
16
C
14 |11
89
KRAS:MKI67
ATM:ATR
LGG
10 18
DSCAM:KRAS LAMB3:PLEC
LIHC
4
24
22 74
4
16 |14
10
11
5
20
26
41
5
39
60
PRKCG CACNA1E
11
LUAD
KRA
33
8
1
16 36
7 4
9
-
23
25
43
LUSC
CR
4
19
50
FBXW7 KRAS
D
KIT:TP53
MESO
5
40
COL11A1:KRAS
OV
Gene pairs
8
4 12
10
13
51
13
9
16
10
22
16
KRAS:SVEP1
5
KRAS LYST
P
10
21
27
ABCB1-ASPM MTOR:PIK3CA
PAAD
7
19
4
PCPG
14
13
1
16
6
a
1
A
3
MYT1:TP53
PRAD
ADAM
ABC
5
8
E
9 4
40
PDE4DIP PDGFRA
E
50
GLI3 KRAS
READ
4
28
19
12
4
1
5
B
KRAS:PDE4DIP
la
MAP2 PTPRD
SARC
BRCA1-TP53
SKCM
PRK
2
7
4
ARCE :- DTOOK
STAD
PF
1
23
A
9
10
21
KRAS MUC17
31
14
4
NOYN!
TGCT
ã
1
11
5
WAPENBAN
THCA
PD
2
11
21
ABCBORDIN
THYM
PC
2
24
5
15
ABCE EPDETDIP
IMBRGAZ
UCEC
7
10
12
11
40
GRM8:PPP1RSA
20
10
45 42
MPTEN
UCS
1
32
ERBBZ JPS2
UVM
1
4
14
5
20
10
12
16
32
FCGBP KRAS
12
31
6
ATM;POLE
6
9
31
40
40
ANK3:PDGFRA
GR
43
16
16
45
DICER1-TP53
21
ATM:PRKDC
5
19
10 11
38
ABCB1:ANK3
ATM:PIK3CA
GÌ
2
6
10
22
13
24
44
FASN:TP53
FRMP
4
16
42
PIK3CA PRKDC
1
13 4
12
11
37
MTOR-TP53
FBX
2
13
11
28
ARID1A SPEN
16
13
a
A
31
DSC
11
11
46
KDR-TP53
DID
6
20
11
4
ATR
HDAC9 TP53
CO
34
5
11
14
B
38
H
42
EGFR TP53
12 44
BRCA2:TP53
L
2
11
KRAS:RYR2
À
14
4
12
9.
F
1
ATM:TP53
16
a
PTEN-TP52
AT
2
11
5
à
4
32
1
4
13
14
PIK3CA:PTEN
ABCE
16
e
15 |17
45
KRAS:TP53
APC:TP53
KICH
SARC
LIHC
GBM-
CHOL
OV
CESC
ACC
DLBC
HNSC
UCS
PAAD
ESCA
LUSC.
BLCA-
SKCM
LUAD
READ-
STAD
COAD
UCEC
O
5
10
0 200 400
Frequency
Frequency
Cancer yp pe
Cancer type
Percentage
100
75
8
25
ABCB1:ANK3
ALMS1:KRAS
APC:TP53
ARID1A:SPEN
ATM:ATR
ATM:PIK3CA
ATM:POLE
ATM:PRKDC
COL11A1:KRAS
EGFR:TP53
ERBB2:TP53
FASN:TP53
FLII:PLEC
GLI3:KRAS
HIPK2:TP53
KDR:TP53
KIT:TP53
KRAS:LYST
KRAS:MUC17
KRAS:POLA1
KRAS:RELN
KRAS:TP53
LAMB3:PLEC
MAP2:PTPRD
MTOR:PIK3CA
PDGFRB:TP53
PML:TP53
PRKCG:PRDM9
PTEN:TP53
RB1:TP53
ABCB1:ASPM
ABCB1:PDE4DIP
ABCB1:PTPRD
ADAMTS18:KRAS
ANK3:PDGFRA
ARHGEF2:TP53
ATM:BRCAZ
ATM:NLRP2
ATM:PTEN
ATM:TP53
ATR:TP53
BRCA1:TP53
BRCA2:TP53
CDKN2A:TP53
CHD7:KRAS
DICER1:TP53
DIDO1:KRAS
DSCAM:KRAS
DSP:TJP1
FBXW7:KRAS
FCGBP:KRAS
FRMPD1:KRAS
GLI2:PIK3CA
GRM8:KRAS
GRM8:PPP1R3A
HDAC9:TP53
KALRN:KRAS
KRAS:MKI6T
KRAS:NRXN1
KRAS:PDE4DIP
KRAS:RYR2
KRAS:SCN4A
KRAS:SVEP1
MET:TP53
MTOR: TP53
MYT1:TP53
PCDHA6:PCDHA9
PDE4DIP:PDGFRA
PIK3CA:PRKDC
PIK3CA:PTEN
PRDM9:PRKCG
PRKCB:SONJA
PRKCG:CACNA1E
PRKDC:RB1CC1
ROS1:TP53
3. ’ Flank UTR
Frame_Shift_Del
Gene pairs
’ Flank
Frame_Shift_Ins In Frame Del
Intron
Missense_Mutation
Silent
5’ UTR
In_Frame_Ins
Nonsense_Mutation
Splice_Site
Translation_Start_Site
B
DR
Freq
A
0.25
0.00
-0.26
UCS-
0
UCEC-
CESC
X
*
*
*
*
COAD-
*
*
SKCM-
*
*
*
K
*
Cancer type
Ov-
*
*
SARC
*
**
LUSC-
* *
*
*
ESCA-
*
HNSC
*
*
*
*
GBM-
*
*
**
LIHC-
*
*
PAAD-
*
BLCA-
*
*
*
LUAD
20 40
ART
Thapsigar
Freq
30
GO
S-Trityl
AZT
V
OGF
GS
PD-
Z-
C
Drug (n = 138)
AK
Gene count 20 15 10
0.08
PIK3CA:PRKDC in BLCA
PIK3CA:PRKDC in BLCA
6.95 |22.98
p = 0.035
p = 0.040
0
-log10(FDR)
7.48
B
17.858.6
adder r
3
Choline
Density
02 0.04 0.06
Probability of survival
roteoglyc
Prostate
Endometna
1
PI3K-AK
ErbB
0.4
Focal
8.39
athway
Pancreatic
.91 12.49 28.99
0.02
Gap
Observed
(45)
Observed
2
-MM
icTORNA
0.00
(40)
-MM
MW
19.3-
-ww
p = 0.0062
WW
p = 0.0226
20
30
40
50
60
37 25.24
20
30
40
50
0
1000
2000
1000
2000
3000
9.37
22.12
6.75
Number of significant gene pairs in COAD Number of significant gene pairs in COAD (survival analysis of 2 groups)
00 3000 4000 50000
4000 5000
Days
Days
KEGG pathway
(survival analysis of 3 groups)
Based on 2 groups
Based on 3 groups
A
810-
CHOL
log2FC
GBM-
3
LUSC
@UCEC
2
LUAD
1
BRCA
0
-
KICH
-1
BLCA KIRC KIRP
-2
-3
U LIHC
O ESCA
exp
STAD
COAD
down
READ
stable
HNSC
up
THCA
PRAD
Invovled gene (n = 429)
0 200 Freq
B
Probability of survival
1.0
PTGS1:WNT5A in KIRC
PTGS1:WNT5A in KIRC
p < 0.001
p = 0.012
0.009
0.8
Density
0.010
0.006
0.2 0.4 0.6
0.005
Observed (1,197)
0.003
Observed (1,441)
AA
0.000
0.000
0.0
-AA
p = 0.0002
- AN
p = 0.0007
-NN
-NN
1000
1050
1100
1150
1200
1300
1400
0
1000
2000
3000
4000
0
1000
2000
3000
4000
Number of significant gene pairs in COAD (survival analysis of 2 groups)
Number of significant gene pairs in COAD (survival analysis of 3 groups)
Days
Based on 2 groups
Days
Based on 3 groups
C
D
log2FC
EXO1
DTL
PKMYT1
KIF14
ASPM
5
-4.22
4.22
CDC20
log_FC
WDR62
CEP55
1.5
NEK2
0
E2F1
-1.5
FOXM1
-5
ECT2
TPX2
HMMR
UBE2C
THCA
READ
PRAD
HNSC
COAD
STAD
KICH
KIRC
ESCA
BLCA
KIRP
BRCA
LUAD
LIHC
UCEC
GBM
CHOL
LUSC
KIF20A
FOSB
PTTG1
MCM10
BUB1B
EGR1
HIST3H2A
ARHGAP11A
PBK
BIRC5
RRM2
Frequency 0 5 10 15
Cancer type
CCNB2
MELK
CHTF18
AURKA
NCAPH
NUSAP1
E
DEPDC1
SORBS1
8485M
CDC6
ALB
MA
PLK1
KIF2C
AURKB
DLGAPS
Ny
RAD
GINS1
MYBLE BUB1
CYP3A4
RAD51AP1
TOP2A
KIT
ARHGA
9
BN TT
200
Frequency
15
1-5
>50
201
TUBBAA
GTSE
NCAPG
SIE
PE
10
NRG2
S
GINS2
CDC45
CENPF
Frequency
150
5
CENPE
6.
-10
31-50
CCNA2
NDCBD
DE!
EE
COCA3
PGR
Gene
0
UBE2T
TRIP 13
NUF2
100
CDK1
Nº
11-15
16-20
MAD2L1
KIFC1
50
2
21-30
82
CALML3
EF
0
Interacted number
di
ACTG2
0
PV
FOX 13
PDIA2
A
A
GAP1
KIFAR
Abnormal expression in cancers
FOSE
PTTO
ARHG
BEBDC
AKIFC
50
E
0
5
RADS
I
a
MA
CH
CALI
Gene
72.59% of partners were up-regulated, respectively), but most mutated genes (>80%) showed normal expression patterns (Fig. 5B and C). Additionally, 30 genes were simultaneously
detected as the first and second genes in different pairs, but rela- tive expression patterns still showed the same expression trends for mutated genes and their partners. Although paired genes were
A
2.0
Mutation level
First_mutation Second_mutation
1.5-
1.0-
0.5-
0.0-
Expression level
0.0-
0.5-
1.0-
1.5-
First_abnormal Second_abnormal
2.0
377 candidate gene pairs
B
Only mutation in first gene
C
Only mutation in sceond gene
First gene (mutation)
300
First gene
200
200
normal
-log padj
150
normal
down
up
down
277 (80.99%)
-log padj
up
392 (72.59%)
100
100
50
0
padj = 0.05
0
padj = 0.05
-6
-3
-1.5
0
1.5
3
-5
-1.5
0
1.5
5
300
Second gene
100
Second gene (mutation)
normal
75
200
-log10padj
up
normal
down
391 (74.90%)
down
-log padj
50
217 (80.37%)
up
100
25
…
0
padj = 0.05
0
padj = 0.05
-8
-4
-1.5
0
1.5
4
8
-5.0
-2.5 -1.5
0.0
1.5
2.5
5.0
log2FC
log2FC
D
Only mutation in first gene
Only mutation in sceond gene
log_baseMean
15
15
10
10
5
5
0
0
down
normal
up
down
normal
up
down
normal
up
down
normal
up
First gene
Second gene
First gene
Second gene
A
B
UCEC
L. Guo, S. LI, B. Qian et al.
0
SKCMComputational and Structural Biotechnology Journal 18 12020) 3243-3254 Structural
COAD-
READ-
1
UCS-
LUAD
2
STAD
Degree
3
40
BLCA-
LUSC
4
CESC.
PAAD-
8
Cancer type
ESCA-
Cancer
freq
6
HNSC.
CHOL
25
10
ACC-
DLBC-
15
10
GBM-
5
OV-
THYM-
10
KICH-
LIHC-
5
UVM-
BRCA-
LGG-
SARC-
MESO-
KIRP.
30
PRAD-
LAML-
TGCT-
KIRC-
PCPG.
THCA
20
PRKCB PRKCG
ERCC6L
PLK11
COL1A1
POLE
MTOR
TOP2A
KIT
KRAS
ERBB2
PDGFRB
KDR
ROS1
MET
RB11
NLRP2
PDGFRA
BRCA2
BRCA1
MAP2
ABCB1
PRKDC
HDAC9
PTEN
EGFR
ASPM
CENPE
ATR
ATM
FBXW7
ARID1A
TP53
Frequency
Gene
C
D
CCNE1
250
CDT1
RAD51
20
(
9
239
90
200
60
122
-
7
30
120
TP53
CDK1
ATM
PRKDC
Frequency
Frequency
Frequency
189
0
113
174
CDC25C
150
N
68
6
13
19
11
119
130
N
TP53
EGFR
4
5
15
XRCC2
EGFR
100
Target number
108
E2F1
OATM
FBXW7
85
92
50
57
64
66
44
12
ARID1A
PRKDC
11
Mutated gene
AURKA
0
11
CDC25C
CDK1
CCNE1
CDT1
XRCC2
RAD51
AURKA
PRKDC
E2F1
TP53
ATM
EGFR
ARID1A
PIF1
TROAP
FBXW7
Gene
E
F
miR-145-5p
Frequency
QNAO
cell cycle
oocyte meiosis
8
MR-139-5p
PRKDC
down
normal
up
EGFR
pathways
p53 signaling pathway
MR-1-3p
RAD51
in cancer
15-
L ==
AURKA
Frequency
et-7c-5p
prostate cancer
pancreatic cancer
ATM
XRCC2
miR-290-3p
CGC
Oncogene
-TSG
actional
essential
drug
MR-424-5p
MUR-486-5p
miR-143-3p
miR-308-5p
CCNE1
5.
muR-378a-3p
CDC25C
COT1
CDK1.
miR-101-3p
CDT1-
0
,
L
%
1
I
7
COK1
1
E2F1
O
L IL
let-Za-5p let-79
1
U
T
O
2
U
T
1
9
let-77-56
let-7i-5P
2
U y
T
A
L
K
A
N
D 5
OD
y
A
X
I
7
2
1
X
3
HAI
JE
6
3
V
2
S
Z
1
EL
S
I
CCNE1
miR-126
mR-379-5p
307
miB-133a
Gene
PRKDC
₹-1
40
5b.
a
b
4
1
A
C
I
A
50
542
674
A
0 40
48a
7-3
N
9
97
204
A
230
34c
45.
m R-10b
N
195a
miR-374b
5
U
MIR-29c-
A
TP53
XRCC2
S
8
miR-2
O
IR-2
O CI MIR- V C
S
A
D
miR-2
miR-148
miR-1
B
-50
5
d
4
H
R-37ª
let -!
R-1
F
2 R- O
miR-199a-5p
AURKA-
R
EEE EEEE
EEEEE
E
R
E E
2
E
S
ELE E
Y
miR-
EE 1
miF
m
mik
E
mil
mik
I
m
miF
RAD51
E2F1
CDC25C
TP53
Related miRNAs
miR-1256-5p
ATM
EGFR
screened for up-regulation, expression trends of mutated genes were not considered during the screening process. These mutated genes showed diverse expression levels in various tissues, and were only rarely dysregulated in some cancer types (Fig. 5B-D), although they were sometimes enriched in some cancer types.
Based on the 91 gene pairs of 78 genes (Table S4), 73.08% showed one or two interactions (46 genes had one interaction and 11 genes had two) (Fig. 6A). KRAS was found to have 25 inter- actions, RAD51 to have 10, and BRCA1 and XRCC2 to have eight each. KRAS has been characterized as a cancer-related gene with potential importance for future cancer treatment [44-46], while RAD51 and XRCC3 polymorphisms may be associated with an increased risk of prostate cancer [47].
To understand potential regulatory patterns of gene pairs con- taining higher mutation frequencies with small non-coding RNAs, we performed an in-depth analysis of 14 gene pairs involving 16 genes (Fig. 6B and Table S5). Of these, TP53 was found to have higher mutation frequencies in 19 cancer types, and five interac- tions with other validated genes (Fig. 6C). Expect for two gene pairs, other interactions showed a network with potential interac- tions among 12 genes. These interactions were further analyzed with respect to miRNAs.
3.6. The regulatory role of small RNAs in synthetic lethal interactions
miRNAs have been widely studied because of their crucial neg- ative regulatory roles in mRNA expression process. Whether the small RNAs also contribute to paired genes with synthetic lethality via coding-non-coding RNA regulatory network? To understand the potential roles of these small RNAs in synthetic lethal interac- tions, related interacting miRNAs for each gene were identified based on biological relationships. Each gene was shown to be reg- ulated by multiple miRNAs, and many miRNAs bound to several mRNA sites (Fig. 6D). These multiple miRNA-mRNA interactions suggested a complex regulatory network of diverse RNAs.
miRNA expression analysis was undertaken according to poten- tial miRNA-mRNA interactions. Because of the existence of multi- ple isomiRs at miRNA loci, we used the most dominant isomiR sequence to analyze detailed expression patterns for each locus. miRNAs were shown to have diverse expression across different tissues, indicating their varied spatiotemporal expression. Because most genes were up-regulated in our analysis (Fig. 6C), a series of miRNAs were identified to construct an miRNA-mRNA network if they were down-regulated in at least four cancer types (Fig. 6E). Thus, we constructed an miRNA-mRNA interaction network (Fig. 6F) showing possible interactions among different RNAs, which may influence related biological pathways.
In this network, we found that many genes contributed to mul- tiple KEGG pathways (Fig. 6F), especially involving the cell cycle (seven of 12 genes), cancer (five of 12 genes), oocyte meiosis, and the p53 signaling pathway. These KEGG pathways are impor- tant in the occurrence and development of cancers, suggesting that the genes have a key role in tumorigenesis. More importantly, many genes were also found to have a close association with the hallmark of cancer, especially evading apoptosis, genome instabil- ity, and mutation. Many were also identified as genes with partic- ular characteristics in tumorigenesis (Fig. 6F). Specifically, EGFR is a widely studied oncogene with a potential role in cancer therapeu- tics [48], six are core genes (AURKA, CDK1, CDT1, PRKDC, RAD51, and XRCC2), six are potential drug targets, and five were identified as drug actionable genes. These potential roles strongly indicated that the genes make direct or indirect contributions to pathology and that synthetic lethal interactions among them will provide impor- tant data for anticancer therapeutic targets.
4. Discussion
Genetic robustness or genetic buffering can contribute to the phenomenon of synthetic lethality, especially because functional genetic redundancy is widespread in many organisms [49,50], typically including the presence of two alleles [51]. Synthetic lethality occurs when the silencing of two genes leads to cell death while silencing of either gene alone does not result in a severe phenotype. It is a possible means of cancer drug target dis- covery [52] and personalized cancer medicine [53] that may be a better approach to specifically kill cancer cells than current treatments.
According to the potential correlations between paired genes with synthetic lethality, we thought that these interacted genes may have complex correlations at different molecular levels. In this study, to understand the potential relationships of interacting genes, especially based on different data sources, we performed a systematic analysis of synthetic lethality between yeast and human data. According to validated gene pairs in yeast, a series of candidate pairs are firstly collected based on evolutionary con- servation. However, further analyses from mutation and expres- sion levels filter many predicted gene pairs, and most remained pairs are human validated or predicted genes. These results impli- cate that predicted synthetic lethal interactions from yeast may not show significant associations via an integrative analysis of multiple molecular levels, while human synthetic lethal interac- tions are prone to be screened to perform in-depth analysis. Indeed, this result is not strange, because predicted gene pairs from yeast are well-conserved genes. These ancient genes may play an important biological role in multiple basic biological processes, implicating that they are very stable than other mutated or abnor- mally expressed genes. Additional screening of candidate gene pairs based on one gene having higher mutation frequencies iden- tified partner gene up-regulated are performed further in-depth analysis. These collected gene pairs contain many genes associated with tumorigenesis (Fig. 6), such as core essential genes, genes of CGC and actionable genes, implicating their possible roles as potential drug targets in cancer treatment. Indeed, genes in the col- lected candidate synthetic lethal interactions may be potential drug target in cancer treatment, and further study based on syn- thetic lethality should be performed to search potential combined medicines.
Furthermore, except for involved genetic interactions, the small RNAs, also play a role in this RNA network. These miRNAs nega- tively regulate these genes directly or indirectly (Fig. 6), and the widespread interactions between miRNAs and mRNAs may con- tribute to gene interactions via coding-non-coding RNA regulatory network. It may be a way to understand synthetic lethal interac- tions via the small regulatory ncRNAs, and the dynamic and popu- lar miRNA:mRNA interactions in vivo will provide more references for studies on synthetic lethality. However, although miRNA:mRNA has been widely studied as an important regulatory patterns between ncRNA and mRNA, multiple isomiRs in miRNA locus should be not ignored. Herein, we only consider the most domi- nant isomiR to perform the relevant analysis, but indeed other iso- miRs are also unexpectedly dominantly expressed. Further studies should focus on the potential roles of multiple isomiRs in synthetic lethal interactions, especially for from the coding-non-coding RNA regulatory network.
Taken together, to understand their potential distributions and relationships, our study analyzes candidate synthetic lethal inter- actions from different sources across molecular levels in diverse cancer types, and then screens a series of gene pairs to identify related regulatory miRNAs. Some gene pairs have important roles in tumorigenesis and potential prognostic value for cancer
treatment. Furthermore, interactions among diverse RNAs compli- cate synthetic lethal interactions, which could contribute to the application of synthetic lethality to personalized anticancer thera- peutics. Further systematic study should be performed based on more candidate data to reveal the potential application in future anticancer therapeutics.
Author contributions
Li Guo: project design, data analyses, manuscript writing. Ting- ming Liang: project design, data analyses, manuscript writing. Sun- jing Li: data analyses. Bowen Qian: data analyses. Youquan Wang: data analyses. Rui Duan: data analyses. Wenwen Jiang: data anal- yses. Yihao Kang: data analyses. Yuyang Dou: data analyses. Guo- wei Yang: data analyses. Lulu Shen: data analyses. Jun Wang: data analyses.
Declaration of Competing Interest
The authors declare that they have no known competing finan- cial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This work was supported by the National Natural Science Foun- dation of China (No. 61771251), the key project of social develop- ment in Jiangsu Province (No. BE2016773), the National Natural Science Foundation of Jiangsu (No. BK20171443), the Nanjing University of Posts and Telecommunications Science Foundation (NUPTSF, No. NY220041), the Qinglan Project in Jiangsu Province, Achievements Incubation Project of the Changzhou Institute of Innovation and Development of Nanjing Normal University (Z201801F06), and the Priority Academic Program Development of Jiangsu Higher Education Institution (PAPD).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.csbj.2020.10.015.
References
[1] Kunjachan S, Rychlik B, Storm G, Kiessling F, Lammers T. Multidrug resistance: physiological principles and nanomedical solutions. Adv Drug Deliv Rev 2013;65:1852-65.
[2] Raz S, Stark M, Assaraf YG. Folylpoly-gamma-glutamate synthetase: a key determinant of folate homeostasis and antifolate resistance in cancer. Drug Resist Updat 2016;28:43-64.
[3] Dobzhansky T. Genetics of natural populations; recombination and variability in populations of Drosophila pseudoobscura. Genetics 1946;31:269-90.
[4] Lucchesi JC. Synthetic lethality and semi-lethality among functionally related mutants of Drosophila melanfgaster. Genetics 1968;59:37-44.
[5] Kaiser CA, Schekman R. Distinct sets of SEC genes govern transport vesicle formation and fusion early in the secretory pathway. Cell 1990;61:723-33.
[6] Hennessy KM, Lee A, Chen E, Botstein D. A group of interacting yeast DNA replication genes. Genes Dev 1991;5:958-69.
[7] O’Neil NJ, Bailey ML, Hieter P. Synthetic lethality and cancer. Nat Rev Genet 2017;18:613-23.
[8] Bryant HE, Schultz N, Thomas HD, Parker KM, Flower D, et al. Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP-ribose) polymerase. Nature 2005;434:913-7.
[9] Roemer T, Boone C. Systems-level antimicrobial drug and drug synergy discovery. Nat Chem Biol 2013;9:222-31.
[10] Pfister SX, Ashworth A. Marked for death: targeting epigenetic changes in cancer. Nat Rev Drug Discovery 2017;16:241-63.
[11] Hartwell LH, Szankasi P, Roberts CJ, Murray AW, Friend SH. Integrating genetic approaches into the discovery of anticancer drugs. Science 1997;278:1064-8.
[12] Blomen VA, Majek P, Jae LT, Bigenzahn JW, Nieuwenhuis J, et al. Gene essentiality and synthetic lethality in haploid human cells. Science 2015;350:1092-6.
[13] Ooi SL, Pan X, Peyser BD, Ye P, Meluh PB, et al. Global synthetic-lethality analysis and yeast functional profiling. Trends Genet 2006;22:56-63.
[14] Srivas R, Shen JP, Yang CC, Sun SM, Li J, et al. A network of conserved synthetic lethal interactions for exploration of precision cancer therapy. Mol Cell 2016;63:514-25.
[15] Reid RJ, Du X, Sunjevaric I, Rayannavar V, Dittmar J, et al. A synthetic dosage lethal genetic interaction between CKS1B and PLK1 is conserved in yeast and human cancer cells. Genetics 2016;204:807-19.
[16] Kirzinger MWB, Vizeacoumar FS, Haave B, Gonzalez-Lopez C, Bonham K, et al. Humanized yeast genetic interaction mapping predicts synthetic lethal interactions of FBXW7 in breast cancer. BMC Med Genomics 2019;12:112.
[17] Neilsen CT, Goodall GJ, Bracken CP. IsomiRs-the overlooked repertoire in the dynamic microRNAome. Trends Genet 2012;28:544-9.
[18] Tan GC, Chan E, Molnar A, Sarkar R, Alexieva D, et al. 5’ isomiR variation is of functional and evolutionary importance. Nucleic Acids Res 2014;42:9424-35.
[19] Guo L, Liang T. MicroRNAs and their variants in an RNA world: implications for complex interactions and diverse roles in an RNA regulatory network. Brief Bioinform 2018;19:245-53.
[20] Telonis AG, Magee R, Loher P, Chervoneva I, Londin E, et al. Knowledge about the presence or absence of miRNA isoforms (isomiRs) can successfully discriminate amongst 32 TCGA cancer types. Nucleic Acids Res 2017;45:2973-85.
[21] Costanzo M, VanderSluis B, Koch EN, Baryshnikova A, Pons C, et al. A global genetic interaction network maps a wiring diagram of cellular function. Science 2016;353.
[22] Berglund AC, Sjolund E, Ostlund G, Sonnhammer EL. InParanoid 6: eukaryotic ortholog clusters with inparalogs. Nucleic Acids Res 2008;36:D263-6.
[23] Chen S, Zhang YE, Long M. New genes in Drosophila quickly become essential. Science 2010;330:1682-5.
[24] Guo J, Liu H, Zheng J. SynLethDB: synthetic lethality database toward discovery of selective and sensitive anticancer drug targets. Nucleic Acids Res 2016;44: D1011-7.
[25] Colaprico A, Silva TC, Olsen C, Garofano L, Cava C, et al. TCGAbiolinks: an R/ Bioconductor package for integrative analysis of TCGA data. Nucleic Acids Res 2016;44:e71.
[26] Yang W, Soares J, Greninger P, Edelman EJ, Lightfoot H, et al. Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells. Nucleic Acids Res 2013;41:D955-61.
[27] Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 2009;4:44-57.
[28] Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome- wide expression profiles. PNAS 2005;102:15545-50.
[29] Futreal PA, Coin L, Marshall M, Down T, Hubbard T, et al. A census of human cancer genes. Nat Rev Cancer 2004;4:177-83.
[30] Hart T, Chandrashekhar M, Aregger M, Steinhart Z, Brown KR, et al. High- Resolution CRISPR screens reveal fitness genes and genotype-specific cancer liabilities. Cell 2015;163.
[31] Wang T, Birsoy K, Hughes NW, Krupczak KM, Post Y, et al. Identification and characterization of essential genes in the human genome. Science 2015;350:1096-101.
[32] Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz Jr LA, et al. Cancer genome landscapes. Science 2013;339:1546-58.
[33] Li J, Han L, Roebuck P, Diao L, Liu L, et al. TANRIC: an interactive open platform to explore the function of IncRNAs in cancer. Cancer Res 2015;75:3728-37.
[34] Esquela-Kerscher A, Slack FJ. Oncomirs - microRNAs with a role in cancer. Nat Rev Cancer 2006;6:259-69.
[35] Calin GA, Croce CM. MicroRNA signatures in human cancers. Nat Rev Cancer 2006;6:857-66.
[36] Li JH, Liu S, Zhou H, Qu LH, Yang JH. starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scale CLIP- Seq data. Nucleic Acids Res 2014;42:D92-7.
[37] Adamcsek B, Palla G, Farkas IJ, Derenyi I, Vicsek T. CFinder: locating cliques and overlapping modules in biological networks. Bioinformatics 2006;22:1021-3.
[38] Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15.
[39] Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, et al. Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res 2003;13:2498-504.
[40] Hanggi K, Ruffell B. Oncogenic KRAS drives immune suppression in colorectal cancer. Cancer Cell 2019;35:535-7.
[41] Hong BJ, Park WY, Kim HR, Moon JW, Lee HY, et al. Oncogenic KRAS sensitizes lung adenocarcinoma to GSK-J4-induced metabolic and oxidative stress. Cancer Res 2019;79:5849-59.
[42] Liu P, Wang Y, Li X. Targeting the untargetable KRAS in cancer therapy. Acta Pharm Sin B 2019;9:871-9.
[43] Jung J, Cho KJ, Naji AK, Clemons KN, Wong CO, et al. HRAS-driven cancer cells are vulnerable to TRPML1 inhibition. EMBO Rep 2019;20:e46685.
[44] Saliani M, Jalal R, Ahmadian MR. From basic researches to new achievements in therapeutic strategies of KRAS-driven cancers. Cancer Biol Med 2019;16:435-61.
[45] Shao YT, Ma L, Zhang TH, Xu TR, Ye YC, et al. The application of the RNA interference technologies for KRAS: current status, future perspective and associated challenges. Curr Top Med Chem 2019;19:2143-57.
L. Guo, S. Li, B. Qian et al.
[46] Aguirre AJ, Hahn WC. Synthetic lethal vulnerabilities in KRAS-mutant cancers. Cold Spring Harb Perspect Med 2018;8:a031518.
[47] Nowacka-Zawisza M, Raszkiewicz A, Kwasiborski T, Forma E, Brys M, et al. RAD51 and XRCC3 polymorphisms are associated with increased risk of prostate cancer. J Oncol 2019;2019:2976373.
[48] Wu M, Zhang P. EGFR-mediated autophagy in tumourigenesis and therapeutic resistance. Cancer Lett 2020;469:207-16.
[49] Tautz D. Redundancies, development and the Flow of Information. BioEssays 1992;14:263-6.
[50] Wilkins AS. Canalization: a molecular genetic perspective. BioEssays 1997;19:257-62.
[51] Hartman JL, Garvik B, Hartwell L. Cell biology - Principles for the buffering of genetic variation. Science 2001;291:1001-4.
[52] Huang A, Garraway LA, Ashworth A, Weber B. Synthetic lethality as an engine for cancer drug target discovery. Nat Rev Drug Discov 2020; 19:23-38.
[53] Jariyal H, Weinberg F, Achreja A, Nagarath D, Srivastava A. Synthetic lethality: a step forward for personalized medicine in cancer. Drug Discov Today 2019;25:305-20.