Gene Expression Meta-Analysis of Colon Rectal Cancer Tumour Cells Reveals Genes in Association with Tumorigenesis

Rutvi Vaja^*
*Correspondence: Rutvi Vaja, Department of Biomedical Sciences and Research, School of Science, Navrachana University, Vadodara-390021, Gujarat, India, Email:

Keywords

Glandular tumours • CRC • GSEA • Meta-analysis • Gene expression

Introduction

As per the statistics of World Health Organization (WHO), Colorectal Cancer (CRC) is the third most prevalent cancer in the world along with fourth leading cause of cancer related deaths [1]. Colon and rectal cancers account for most of the glandular malignancies with the incidences increasing with age [1]. Highly penetrant, autosomal dominantly or recessively inherited tumour predispositions cause about 5% of all colorectal cancers [2]. More than 945 000 people are diagnosed with colorectal cancer each year, with roughly 492 000 patients dying [3]. This type of cancer occurs infrequently, often as a result of genetic cancer syndromes or inflammatory bowel illnesses [3]. As per the GLOBOCAN database documented 1.8 million newly diagnosed cases of CRC and 861,600 cases of CRC-related mortality over the world [4]. CRC is a highly diverse disease caused mostly by interactions between genetic changes and environmental variables [5]. Despite advances in diagnosis and treatment, the survival rate of CRC patients has remained unchanged over the previous two decades, with more than half of patients having regional or distant metastasis at the time of diagnosis [6-8]. Several genes and cellular signalling pathways, including RACK1 (receptor for activated C kinase 1) and long non- coding RNA breast cancer anti-estrogen resistance 4 (lncRNA BCAR4), have been implicated in the formation and progression of CRC [6]. RACK1 expression, for example, has been shown to be considerably upregulated in CRC tissues when compared to adjacent normal tissues [6]. Despite this comprehensive and meticulous research to find novel targets for CRC management, a comprehensive description of the critical key genes and signalling pathways implicated in CRC is lacking, to the best of our knowledge [4-6].

A high-throughput method for detecting mRNA expression in tissues, gene microarray profile analysis, is becoming a more promising tool in medical oncology. An enhanced understanding of the molecular pathogenesis of many cancer types can be gained by analysing differential gene expression between tumour tissues and normal control tissues, allowing for the identification of prospective target genes and signalling pathways for precision medicine [7]. In earlier decades, microarray technology was employed in various researches on gene expression profiles in cancer, but only one study focused on CRC [8]. Apart from these studies, the comparative analysis of Differentially Expressed Genes (DEGs) remains relatively limited [9]. Furthermore, more research is needed to identify meaningful leading edge gene profiles for distinguishing CRC from normal tissues. In addition, the relationships among the DEGs should be clarified, along with the interaction networks and critical biological signalling pathways affected.

Many data mining analyses of mRNA, microRNA, long non-coding RNA, and DNA methylation have been performed on human cancers, particularly colon cancer, over the last few decades. A complete understanding about the molecular changes associated with CRC Tumor formation might help in assisting the development of new therapeutic interventions. Several studies have been done to elucidate molecular changes associated with CRC by examining gene expression changes in CRC- tumour forming cell cultures [10- 12]. There are some studies which provide a clinical reference for predicting the survival probability of patients with different clinical subtypes [13]. Other studies generated Protein-protein Interaction (PPI) networks, and centrality analysis as well was performed in order to identify the crucial genes that were potentially involved in the development of CRC [11]. Several bioinformatics tools and strategies have aided in the exploration of molecular mechanisms of tumour pathogenesis and provided clues for better knowledge of related malignancies by identifying early biomarkers and potential therapeutic targets of tumours. Colon cancer is a multifactorial disease caused by a variety of factors including genetic, environmental, and lifestyle influences, but the pathogenesis of the disease is yet unknown. Exploring and analysing colon cancer's molecular basis and important genes is critical to improve colon cancer prevention and treatment.

Previous work using a GSEA-based meta-analysis approach successfully identified known and novel genes associated with severe acute respiratory syndrome (SARS) infection through differential gene expression comparison between mRNA expression datasets. Therefore, this paper applied the same GSEA-based approach to analyze mRNA expression data of tumor and normal CRC tissue derived from human colonic biopsy samples by defining and comparing gene expression signatures (i.e., gene lists ranked by differential expression) [14,15].

In this particular study, we have used expression profiling array datasets to identify and verify gene expression changes associated with CRC tumour formation to elucidate molecular changes associated with the process of tumourogenesis in CRC pathogenesis. Gene expression changes associated with CRC tumour formation then were compared to changes resulting from other colonic tumour forming diseases like Ulcerative colitis, adenoma dysplasia and Hyperplastic Polyp, to examine the common leading edge genes in CRC that invoked changes and played a potential role in CRC tumour formation. Finally we verified the gene signatures in several other colonic tumour forming diseases such as Ulcerative coilitis, adenoma dyslapsia and Hyperplastic Polyp. A meta-analysis was performed lastly to find the common shared leading edge genes across all different conditions of CRC. The gene expression changes identified with across these several conditions holds the potential to actually improve the treatments targeted for CRC and its therapeutic intervention.

Methods

mRNA expression resources

To identify gene expression changes associated with CRC tumour forming cells, the Gene Expression Omnibus (GEO) repository was searched to find datasets for use in this study (Table 1) [16]. The six independent data series GSE44861, GSE113513, GSE44076 GSE10714, GSE32323 and GSE24514 had CRC tumour and normal tissue samples and hence we started our analysis from here. GSE44861 is an Affymetrix expression data collected from colon cancer patient tissues in which RNA from fresh frozen colon tissues were extracted using Trizol and hybridized to Affymetrix U113A arrays [17]. The platform used for GSE44861 was GPL3921 [HT_HG-U133A] Affymetrix HT Human Genome U133A Array. GSE113513 has samples from 14 colorectal cancer patients who had undergone surgical resection of colorectal cancer where Trizol (Thermo Fisher Scientific) extraction of total RNA was performed [18]. The platform used for GSE113513 was GPL15207 [PrimeView] Affymetrix Human Gene Expression Array. GSE44076 contained Gene expression profiles of paired normal adjacent mucosa and tumor samples from 98 individuals and 50 healthy colon mucosae, which were obtained through Affymetrix Human Genome U219 Arrays [19]. The platform used for GSE44076 was GPL13667 [HG-U219] Affymetrix Human Genome U219 Array. GSE10714 had expression data from human colonic biopsy samples on which Qiagen RN easy Mini extraction of total RNA was performed [20]. The platform used for GSE10714 was GPL570 [HG-U133_Plus_2] Affymetrix

Human Genome U133 Plus 2.0 Array. GSE32323 consisted of gene expression profiles for 17 pairs of cancer and non-cancerous tissues from colorectal cancer patients were measured by Affymetrix HG- U133 plus 2.0 arrays. Here, total RNA was extracted from tissue specimens using Reasy kit (Qiagen, Hilden, Germany) [21]. The platform used for GSE32323 was GPL570 [HG-U133_Plus_2].

Affymetrix Human Genome U133 Plus 2.0 Array. Lastly, GSE24514 had expression data from human MSI colorectal cancer and normal colonic mucosa in which RNA rom fresh frozen tissues was extracted with Trizol reagent (Invitrogen) [22]. The platform used for GSE24514 was GPL96 [HGU133A] Affymetrix Human Genome U133A Array].

The expression data provided by GEO for all datasets were z-scored normalized across all samples within the dataset regardless of tumour cell or treatment type prior to use for analysis. The expression data was cleaned by removing probe identifiers given by GEO where 1) all samples having gene expression z-score of 0, or 2) duplicate identifiers/symbols were identified so only the identifier with the highest coefficient of variation were retained. The soft family data consisted of Entrez Gene ids, Gene symbols and Probe identifiers as well. If a dataset’s GEO-provided platform contained both Ensemble gene IDs and gene symbols for a probe, the GEO-provided platform files to convert between these two prove identifiers were used. For this particular study, we have converted the probe identifiers to gene symbols using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.8 gene conversion tool in order to maintain uniformity [23]. In order to obtain other meta- data information like sample characteristics, sample source, Ensemble ids and GSM numbers, GEO- provided platform files were used.

Defining gene signatures

In order to examine the gene expression changes associated with tumour formation in CRC differential gene expression was measured for samples of interest from each dataset using Welch’s two-sample T test score of cleaned and normalized values. (For eg: tumour samples and normal tissue sample from one dataset formed one signature). Samples with the same origin, regardless of tumour stage, cancer type (metastasis or benign) were combined to form one signature. The resulting lists of genes along with their T-test scores were used to define 13 gene signatures. These signatures are formed from the gene lists ranked from high to low differential gene expression between tumors vs. normal tissue samples. The signatures derived from the same dataset used the same control samples. The gene location where T-score becomes negative (i.e., T-score = 0) and the T- score range for each signature are found in Table 2.

Identification of genes associated with CRC tumour tissues

To identify gene expression changes associated with CRC Tumour tissues, two CRC tumour gene panels were generated (Figure 1). In order to do this, 500 genes from the positive and negative tails from both the GSE44861- derived Tumour vs. Normal and GSE113513-derived Tumour vs. Normal gene signatures were selected and then were used to form four individual query gene sets. GSEA compared each query gene set to the both the entire GSE44861-derived Tumour vs. Normal and GSE113513-derived Tumour vs. Normal gene signatures (reference). For our identification stage, we used the datasets GSE44861 as query first against GSE113513 as reference followed by GSE113513 as query dataset against the whole of GSE44861 Tumour vs. Normal signature. Leading- edge (LE) genes from each of these analyses were examined and shared leading-edge genes were used to define two CRC Tumour vs. Normal gene panels, one panel per tail(Positive CRC and Negative CRC Panels). Pathway enrichment analysis was performed on both CRC tumour gene panels using DAVID.

Verification of CRC tumour gene panels

To verify the CRC Tumour gene panels, GSEA between CRC Tumour gene panels (Pos. CRC and Neg. CRC) and GSE44076-derived and GSE10714-derived Tumour vs. Normal signatures was performed. To assess if results generated from GSEA could be achieved randomly, 1000 gene panels consisting of 175- genes to match the average number of genes in the positive and negative CRC Tumour panels (query signature) were randomly selected from the GPL3927 platform used to define the GSE44861 gene signatures used for gene identification for GSEA against GSE44076-derived and GSE32323- derived Tumour vs. Normal signatures (reference signature). These analysis generated a null distribution of NES(Normalized enrichment score) to which were compared the NES achieved by CRC tumour gene panels for each reference gene signature and count the number of equal or better NES to estimate significance (i.e., null distribution p-value). The bar and whiskers Plot was created and calculated using Excel. Heat maps were generated by a user-friendly, web-based program maps. https://software.broadinstitute.org/ morpheus. Morpheus is software created by Broad Institute Software.

Comparison of CRC tumour gene panels to other types of tumour that are glandular in origin

GSEA was used to compare the identified CRC gene panels and gene signatures derived from samples of one of the six glandular tumour (Ulcerative Colitis, Hyperplastic Polyp, Adenoma with low grade dysplasia, Colorectal cancer – Benign, Colorectal cancer with a chronic disease and colorectal Tumours with microsatellite instability), compared to normal tissue samples was performed to compare gene expression changes across CRC tumour tissues and several other glandular tumours. Following this, random modelling as stated earlier was used to assess if results generated from GSEA could be achieved randomly. Leading-edge genes from each statistically significant (GSEA P- value <0.05), non-random (null distribution P-value < 0.05) GSEA were examined and analysed for common genes.

Gene signature approach identified gene expression changes associated with colorectal cancer tumor tissues in vitro

GSE44861-derived and GSE113513-derived Tumour vs. Normal gene signatures were defined to identify genes associated with response to Tumour tissue of human CRC colonic biopsy samples (Table 2). From both these two Tumour vs. Normal identification signatures, two gene sets were generated containing the 500 most differentially expressed genes from the positive and negative tails of each signature, capturing maximum coverage of the signature that was allowable by GSEA [24]. The T-score values for GSE44861 derived Tumour vs. Normal signature was >10.77 and <-11.89. For GSE113513 derived Tumour vs. Normal gene signature was >12.95 and <-14.0282. To find the similarity between these two signatures, enrichment was first calculated using GSEA between GSE113513- derived Tumour vs. Normal positive or negative tail gene sets (individual query set) and the GSE44861- derived Tumour vs. Normal(reference set) and achieved NES = 3.54 and NES = -3.64 for positive and negative tail query gene sets, respectively, both with a GSEA p-value <0.001. Similarly enrichment was calculated using GSEA between GSE44861- derived Tumour vs. Normal positive or negative tail gene sets (individual query set) and the GSE113513- derived Tumour vs. Normal (reference set) and achieved NES = 3.38 and NES = -4.14 for positive and negative tail query gene sets, respectively, both with a GSEA p-value <0.001. The identified leadingedge genes from GSEA are listed in Supplementary Table 1.

No genes in the positive CRC panel were mentioned in the published reports for GSE44861 and GSE113513, though some panel genes like Kruppel like factor-7(ubiquitous) (KLF-7) and RNA terminal phosphate cyclaselike 1(RCL-1) have reported connections with CRC tumour formation in the literature [25-28]. In the negative CRC gene panel, Keratin 20(ALOX) was found to have previous associations with CRC tumour formation [29]. Taking all the results together, this demonstrated the detection ability of using a GSEAbased approach to gene identification. The rest of the genes in the negative panobinostat panel had no prior association with CRC tumour formation from the published reports for GSE44861 and GSE113513. It can be speculated that genes lacking previously reported associations with CRC tumour formation that were identified here also are associated with human colonic biopsy samples.

To expand our analysis, the cellular roles of CRC gene panels were examined using DAVID to calculate enrichment between each CRC gene panel and pathways in popular known knowledgebase. It was noticed that, when compared to other databases, the GO BP database returned the most significantly enriched pathways (data not mentioned here) hence this discussion was concentrated on GO-BP data to prevent confusion caused by other overlapping pathway and gene inclusion differences across other multiple known knowledgebase.

DAVID identified nine significant GO-BP pathways (EASE score P-value < 0.05) from the positive CRC gene panel and 10 significant pathways from the negative CRC gene panel (Supplementary Table 2). Some significantly enriched pathways have experimentally established associations with CRC tumour formation, such as RNA processing pathway (GO: 0006396, P-value = 0.02), demonstrating the ability of our gene signature approach to detect pathways associated with CRC Tumour formation [25]. Other identified pathways, such as lipid catabolic process (GO: 0016042, P- value=0.04), have no prior associations to CRC Tumour formation. Therefore, it can be speculated that pathways that came out as a result without prior association to CRC Tumours identified here also were involved in CRC.

Enrichment of CRC gene panels and specific CRC panel genes verified in independent datasets

To verify our CRC gene panels, GSEA was used to calculate enrichment between our CRC panels (individual queries) and two verification gene signatures (individual references): GSE32323-derived Tumour vs. Normal and GSE44076-derived Tumour vs. Normal (Table 2). Significant similarity between positive and negative CRC panels and GSE32323-derived Tumour vs. Normal (NES = 2.26 for the positive CRC panel, Figure 2A, and NES = -2.56 for the negative CRC panel, Figure 2B, both GSEA P- value < 0.001) was found. To determine how likely the NES achieved for CRC gene panels would be achieved by random chance, 1000 randomly selected 175-gene panels were generated from the GSE44861-derived Tumour vs. Normal gene signature to match the average size and potential composition of our CRC gene panels. GSEA was then repeated using these randomly generated gene panels (individual queries) and the GSE44076-derived Tumour vs. Normal (reference) to generate a null distribution of NES achieved via a random chance.

From this, random NES ranged from 1.47 to -1.5 was found (data not shown), illustrating that NES achieved by our CRC panels are non-random (null distribution P-value <0.001). Taken together, these results demonstrate that the enrichment achieved from our panobinostat panels was true.

To determine which of our CRC panel genes were verified across all signatures, leading-edge genes identified by GSEA for each verification signature were examined. Leading-edge genes for GSE44076-derived and GSE32323-derived Tumour vs. Normal signatures are listed in Supplementary Tables 3-10. 51 genes from the positive CRC panel and 75 genes from the negative CRC panel were shared between verification signatures. These data together verify our shared leading-edge genes are associated with CRC tumour tissues in human colonic biopsy samples and support the hypothesis that identified genes without previously reported associations are also associated with CRC tumour tissues in human colonic biopsy samples.

Non-random enrichment of CRC tumour gene panels found in other glandular tumours

To expand this study, gene expression changes associated with CRC Tumour were compared to changes observed in other glandular tumours included in GEO Series previously used to examine the common shared leading edge genes between CRC tumour tissues and other glandular tumours. The following eight signatures were examined: Adenoma vs. Normal, Hyperplastic Polyp vs. Normal, Uc vs. Normal, CRCB vs. Normal, CRCD vs. Normal, Tumour M vs. Normal M, Tumour MR vs. Normal MR and MSIT vs. Normal (Table 2). For Adenoma vs. Normal gene signature vs. the positive panel NES = 1.785 and for Negative panel is NES = -2.00 (GSEA P. value <0.001). For Hyperplastic Poly vs. Normal gene signature, when running a query against the Positive panel NES = 1.86 and for negative panel NES = -2.022 m (GSEA P. value <0.001). For Uc vs. Normal gene signature, NES for positive panel = 1.54 and NES for negative panel is -1.87 m (GSEA P. value <0.001). For CRCB vs. Normal gene signature against positive panel query, NES = 0.9 and for negative panel query, NES = -1.7(GSEA P. value <0.001). For CRCD vs. Normal gene signature, the query against Positive panel yielded a NES of 1.866 whereas the NES of negative panel was -2.00 (GSEA P. value <0.001). The Tumour M vs. Normal M gene signature yielded NES of 2.39 for the positive gene panel as the query set and NES of -2.47 (GSEA P. value <0.001). For Tumour MR vs. Normal MR the NES for the positive panel was 2.33 and for the negative panel NES = -2.5 (GSEA P. value <0.001). For gene signature of MSIT vs. Normal the positive panel’s NES score is 2.307 and negative panel’s NES score is -3.12 (GSEA P. value <0.001). Both the positive and negative CRC panel showed significant enrichment score (GSEA P. value <0.001) in all the gene signatures with and were also non-random (null distribution P- values <0.001).

Leading-edge genes found amongst the other glandular tumours

Finally, leading-edge genes from each GSEA across other glandular tumours were examined to identify genes of potential interest. Supplemental Material S Tables 3 through 9 contains leading-edge genes from GSEA with each CRC gene panel across 8 comparison gene signatures (Table 2). Out of the 54 common shared genes in the Positive CRC panel across all the signatures, 48 genes stood out having a statistical significance of (P value <0.001). Whereas out of the 75 common shared genes in the negative CRC panel across all the signatures, 72 genes were of highly statistical significance of (P value <0.001).9 solute carrier family genes solute carrier family 25, member 32 /// solute carrier family 25 member 32(SLC25A32), solute carrier family 22 (extraneuronal monoamine transporter) member 3 (SLC 22 A 3), solute carrier family 25 (carnitine/acylcarnitine translocase) member 20 (SLC25A20),solute carrier family 36 (proton/amino acid symporter) member 1 (SLC36A1), solute carrier family 17 (sodium phosphate) member 4 (SLC17A4),solute carrier family 26 member 3 (SLC26A3), solute carrier family 9 (sodium/hydrogen exchanger)member 2 (SLC9A2), solute carrier family 4 sodium bicarbonate cotransporter member 4 (SLC4A4),and solute carrier family 26 (sulfate transporter), member 2(SLC26A2) from the CRC panel were shared commonly across all the gene signatures. Group of carbonic anhydrase protein coding genes from CRC gene Panel were also commonly shared across all gene signature’s namely carbonic anhydrase XII (CA12), carbonic anhydrase II (CA2), carbonic anhydrase IV (CA4) and carbonic anhydrase I (CA1) (Figure 3).

To rank shared leading-edge genes, T-test P value analysis using Excel- STAT software was used. Out of the 129 Leading edge genes (54 from positive gene panel and 75 from negative gene panel) we wanted to find out the significant leading edge genes. Hence using the parameter: P. Value < 0.001 we applied a P- value T-test in Excel to calculate the P-value. A total of 120 genes (48-Over-expressed and 72 under-expressed) out of the 129 were significantly valid leading edge genes. The volcano plot for the significant leading-edge genes is shown in (Figure 4 and Table 4).

Results and Discussion

Colorectal cancer still stands as a predominant cancer and is the secondand third-most common cancer in women and men, respectively [26]. Despite considerable breakthroughs in treatment, colorectal cancer mortality remains high, with 40–50 percent of patients dying as a result of their illness. Identification of differentially expressed genes associated with CRC itself can contribute to the overall understanding of the molecular changes that drive cellular and molecular changes in CRC tumour development. This improved understanding can potentially contribute to the development of new therapeutic options to improve the prognosis for CRC patients. This work conducted a metaanalysis of gene expression signatures generated from mRNA expression data across CRC tumours and eight other glandular tumours which are colonic in nature, to identify differentially expressed genes associated with CRC tumour formation. Genes that change in response to treatment may contribute to developing treatment resistance long-term.

In this study, out of the 55 genes from the positive CRC gene panel, 54 of them were continuously constant throughout all the gene signatures and for the negative gene CRC panel which had 77 genes, 75 of them were uniform across all the other gene signatures. To find out whether, these genes are statistically significant or not we did a T-test for calculating the p value in excel. Among the genes identified in this study, from the positive CRC gene panel, 54 common shared genes in the Positive CRC panel across all the signatures, 48 genes stood out having a statistical significance of (P value < 0.001). Whereas out of the 75 common shared genes in the negative CRC panel across all the signatures, 72 genes were of highly statistical significance of (P value <0.001). Among the genes identified, studies have shown that nuclear factor (erythroidderived 2)-like 3(NFE2L3) decreases colon cancer cell proliferation in vitro and tumor growth in vivo [30]. Interestingly, sterile alpha motif and leucine zipper containing kinase AZK (ZAK) gene has involved functionalities in lung cancer tumourigenesis process and JNK pathway activation [31]. Methyltransferase like 7A (METTL7A) have shown previous associations with thyroid cancer but not in lung, uterine, ovarian, gastric, esophagus, pancreatic, liver, or colorectal cancers via bioinformatic analysis [32]. Aken together, these results suggested that the GSEA-based meta-analysis approach used here was successful in identifying cancer-related genes with and without CRC tumourigenesis associations.

This work had observable gene detection limitations that may have biological implications. For example, gene expression changes commonly associated with CRC tumourigenesis in genes like MSH2 and MSH6 both on chromosome 2 and MLH1, on chromosome 3 were not found in this study [4,30,33]. Platform variations, both in gene inclusion and primer nucleotide sequence, can substantially impact results generated from this, or any, bioinformatics analysis. While MSH2 and MSH6 were included in the GSE113513 identification dataset, their T-scores were insufficient to make the 500 gene cut-off required of GSEA to maintain statistical accuracy. However, this is an inherent limitation with the usage of GSEA, based approach which can only be overcomes by switching to a non-GSEA based approach. However, if the desired outcome is a prioritized list of potential gene candidates for further laboratory or clinical examination, the GSEA-based approach used here suffices (Tables 3 and 4).

A lack of direct experimental or clinical evidence substantially limited the conclusions drawn from this study of purely bioinformatics comprised work. Follow-up experiments must be done using laboratory techniques, such as Western blotting or qRT-PCR, to confirm top gene candidate predictions would support these conclusions drawn exclusively from mRNA expression data using GSEA based approach. Further analysis examining gene expression data from colonic biopsy samples undergoing any type of treatment for CRC that mimics clinical samples is needed to assess the prediction of results portrayed here. Further, gene expression data directly from CRC patients would be of particular interest to further explore the results generated here. Such an examination of gene expression data from treated and untreated human biopsy colonic CRC tissue samples would be limited due to challenges acquiring samples from tumor location.

Conclusion

This work used mRNA micro-array expression data to predict genes potentially involved in developing CRC tumours examining gene signatures. Through a GSEA-based meta-analysis approach, 54 over- expressed genes, most important ones being SLC25A32, SLC22A3, CA12, CA4, CA1 and SLC4A4, were identified amongst the 54 over-expressed genes as being most associated with CRC tumourigenesis regardless of its association with any other glandular tumour in origin. Overall, this work demonstrated the usefulness of a meta-analysis approach used previously to detect genes associated with SARS infection in identifying genes associated with DIPG treatment through application on mRNA expression data. Also, further laboratory, wet lab experiments and clinical examination into the role these identified gene

expression changes play in developing CRC tumours might improve treatment options available and life expectancy for CRC patients.

Declaration of Interests

The authors declare no conflicts of interest.

References

1. Haraldsdottir, Sigurdis, Hulda M. Einarsdottir, Agnes Smaradottir and Adalsteinn Gunnlaugsson, et al. "Colorectal cancer-review." Laeknabladid 100 (2014): 75-82.

Google Scholar, Crossref

2. Heinimann, Karl. "Hereditary colorectal cancer: Clinics, diagnostics and management." Therapeutische Umschau. Revue Therapeutique 75 (2018): 601-606.

Indexed at, Google Scholar, Crossref

3. Weitz, J., M. Koch, J. Debus and T. Höhler, et al. Colorectal cancer. Lancet 365 (2005):153-165.

Crossref

4. Bray, Freddie, Jacques Ferlay, Isabelle Soerjomataram and Rebecca L. Siegel, et al. "Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries." CA: A Cancer J Clin 68 (2018): 394-424.

Indexed at, Google Scholar, Crossref

5. Andersen, Stig E., Ida B. Andersen, Benny V. Jensen and Per Pfeiffer, et al. "A systematic review of observational studies of trifluridine/tipiracil (TAS-102) for metastatic colorectal cancer." Acta Oncologica 58 (2019): 1149-1157.

Indexed at, Google Scholar, Crossref

6. Jin, Shaoju, Yishuang Mu, Xiaodong Wang and Zhenru Liu, et al. "Overexpressed RACK1 is positively correlated with malignant degree of human colorectal carcinoma." Mol Biol Rep 41 (2014): 3393-3399.

Indexed at, Google Scholar, Crossref

7. Shangguan, H, S.Y. Tan and J.R. Zhang. "Bioinformatics analysis of gene expression profiles in hepatocellular carcinoma." Eur Rev Med Pharmacol Sci 19 (2015): 2054-2061.

Indexed at, Google Scholar

8. Yu, Jiekai, Xiaofen Li, Chenhan Zhong and Dan Li, et al. "High-throughput proteomics integrated with gene microarray for discovery of colorectal cancer potential biomarkers." Oncotarget 7 (2016): 75279.

Indexed at, Google Scholar, Crossref

9. Shen, Xingjie, Meng Yue, Fansheng Meng and Jingyu Zhu, et al. "Microarray analysis of differentially-expressed genes and linker genes associated with the molecular mechanism of colorectal cancer." Oncol Lett 12 (2016): 3250-3258.

Indexed at, Google Scholar, Crossref

10. Behrens, J. "The role of the Wnt signalling pathway in colorectal tumorigenesis." Biochem Soc Transac 33 (2005): 672-675.

Google Scholar, Crossref

11. Worthley, Daniel L. and Barbara A. Leggett. "Colorectal cancer: Molecular features and clinical opportunities." Clin Biochem Rev 31 (2010): 31.

Indexed at, Google Scholar

12. Grady, William M. "Genomic instability and colon cancer." Cancer Metas Rev 23 (2004): 11-27.

Indexed at, Google Scholar, Crossref

13. Fang, Zhengyu, Sumei Xu, Yiwen Xie and Wenxi Yan. "Identification of a prognostic gene signature of colon cancer using integrated bioinformatics analysis." World J Surg Oncol 19 (2021): 1-14.

Indexed at, Google Scholar, Crossref

14. Aran, Veronica, Ana Paula Victorino, Luiz Claudio Thuler and Carlos Gil Ferreira. "Colorectal cancer: Epidemiology, disease mechanisms and interventions to reduce onset and mortality." Clin Colorectal Cancer 15 (2016): 195-203.

Indexed at, Google Scholar, Crossref

15. Park, Amber and Laura K. Harris. "Gene expression meta-analysis reveals interferon-induced genes Associated with SARS infection in lungs." Front Immunol 2 (2021): 2924.

Indexed at, Google Scholar, Crossref

16. Clough, Emily and Tanya Barrett. "The gene expression omnibus database." In StatisticalGenomics, Humana Press, New York, United States.

Indexed at, Google Scholar, Crossref

17. Ryan, Bríd M., Krista A. Zanetti, Ana I. Robles and Aaron J. Schetter, et al. "Germline variation in NCF4, an innate immunity gene, is associated with an increased risk of colorectal cancer." Int J Cancer 134 (2014): 1399-1407.

Indexed at, Google Scholar, Crossref

18. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi
19. Berdiel-Acer, Mireia, Rebeca Sanz-Pamplona, Alexandre Calon and Daniel Cuadras, et al. "Differences between CAFs and their paired NCF from adjacent colonic mucosa reveal functional heterogeneity of CAFs, providing prognostic information." Mol Oncol 8 (2014): 1290-1305.

Indexed at, Google Scholar, Crossref

20. Leiszter, Katalin, Ferenc Sipos, Orsolya Galamb and Tibor Krenacs, et al. "Promoter hypermethylation-related reduced somatostatin production promotes uncontrolled cell proliferation in colorectal cancer." PLoS One 10 (2015): e0118332.

Indexed at, Google Scholar, Crossref

21. Galamb, O, S. Spisák, F. Sipos and K. Tóth, et al. "Reversal of gene expression changes in the colorectal normal-adenoma pathway by NS398 selective COX2 inhibitor." British J Cancer 102 (2010): 765-773.

Indexed at, Google Scholar, Crossref

22. Alhopuro, Pia, Heli Sammalkorpi, Iina Niittymäki and Mia Biström, et al. "Candidate driver genes in microsatellite‐unstable colorectal cancer." Int J Cancer 130 (2012): 1558-1566.

Indexed at, Google Scholar, Crossref

23. Huang, Da Wei, Brad T. Sherman, Xin Zheng and Jun Yang, et al. "Extracting biological meaning from large gene lists with DAVID." Curr Protocols Bioinformat 27 (2009): 13-11.

Indexed at, Google Scholar, Crossref

24. Subramanian, Aravind, Pablo Tamayo, Vamsi K. Mootha and Sayan Mukherjee, et al. "Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles." Proceed Nat Acad Sci 102 (2005): 15545-15550.

Indexed at, Google Scholar, Crossref

25. Oosterwijk, E. and R.J. Gillies. "Targeting ion transport in cancer." Philosoph Transact Royal Soc 369 (2014): 20130107.

Indexed at, Google Scholar, Crossref

26. Kuipers, E.J, W.M. Grady, D. Lieberman and T. Seufferlein, et al. "Colorectal cancer.” Nat Rev Disease Primers 1 (2015): 15065.

Google Scholar, Crossref

27. Bureau, Christophe, Naima Hanoun, Jérôme Torrisani and Jean-Pierre Vinel, et al. "Expression and function of Kruppel like-factors (KLF) in carcinogenesis." Curr Genomics 10 (2009): 353-360.

Indexed at, Google Scholar, Crossref

28. Boland, C. Richard and Ajay Goel. "Microsatellite instability in colorectal cancer." Gastroenterol 138 (2010): 2073-2087.

Indexed at, Google Scholar, Crossref

29. Goodman, Julie E., Elise D. Bowman, Stephen J. Chanock and Anthony J. Alberg, et al. "Arachidonate lipoxygenase (ALOX) and cyclooxygenase (COX) polymorphisms and colon cancer risk." Carcinogenesis 25 (2004): 2467-2472.

Indexed at, Google Scholar, Crossref

30. Bury, Marina, Benjamin Le Calve, Frédéric Lessard and Thomas Dal Maso, et al. "NFE2L3 controls colon cancer cell growth through regulation of DUX4, a CDK1 inhibitor." Cell Rep 29 (2019): 1469-1481.

Indexed at, Google Scholar, Crossref

31. Yang, Jaw‐Ji, Yi‐Ju Lee, Hsin‐Hung Hung and Wei‐Pu Tseng, et al. "ZAK inhibits human lung cancer cell growth via ERK and JNK activation in an AP‐1‐dependent manner." Cancer Sci 101 (2010): 1374-1381.

Indexed at, Google Scholar, Crossref

32. Zhou, Shasha, Yihang Shen, Min Zheng and Linlin Wang, et al. "DNA methylation of METTL7A gene body regulates its transcriptional level in thyroid cancer." Oncotarget 8 (2017): 34652.

Indexed at, Google Scholar, Crossref

33. https://www.ncbi.nlm.nih.gov/books/NBK22218/