Molecular classification of colorectal cancer using the gene expression profile of tumor samples

Abstract

Molecular classifications of colorectal cancer are benefitting cancer research by providing insights into subtype-specific disease prognosis and improved therapeutic interventions. Different conventional DNA markers, such as microsatellite instability, CpG island methylator phenotype, chromosomal instability, and BRAF and KRAS mutations, have been used to classify colorectal cancer patients but have not yet shown promising prognostic values. Here, for the first time, to the best of our knowledge, we show a classification of colorectal cancer tumors from Saudi Arabian patients based on the gene expression profile. An existing method of colorectal cancer subtyping has been applied to the gene expression profile of tumors from Saudi colorectal cancer patients. A survival analysis was done on the predicted colorectal cancer subtypes. In silico functional analyses were conducted on the gene signature used for the subtype prediction. The predicted subtypes showed a distinct but statistically insignificant overall survival distribution (log-rank test, P = 0.069). A comparison of the predicted subtypes in Saudi colorectal cancer patients with that of French patients showed significant dissimilarity in the two populations (Chi-square test, P = 0.0091). Functional analyses of the gene signatures used for subtyping suggest their association with “cancer” and “gastrointestinal diseases.” Most of the signature genes were found differentially expressed in colorectal cancer tumors compared to adjacent normal tissues. This classification framework might facilitate the treatment of colorectal cancer patients.

Impact statement: Colorectal cancer is a heterogeneous disease and subtyping could be useful in implementing precision medicine approach. In this report, we identified molecular subtypes in relatively less studied CRC patients from Saudi Arabia using the prediction model developed on the French population. The predicted subtypes showed distinct overall survival among the six subtypes. Chi-square results exhibited the dissimilarity between French and Saudi colorectal cancer patient population in terms of subtype distribution (P value = 0.0091). Gene signature (57 genes) used for subtyping was found to be functionally relevant as evident from the pathway analyses. These genes were found to be associated with gastrointestinal disease and cancer. Genes used for subtyping were found to be differentially expressed in Saudi colorectal cancer patient samples when compared with their own normal tissue. Taken together, this study supports a classification method for CRC patients by using patient samples from a different geographical region.

Keywords: Colorectal cancer; exon ST; gene expression; molecular classification; molecular subtyping; survival analysis; tumor classification.

Figure 1.

Overall analysis methodolgy adopted in the current study. (A color version of this figure is available in the online journal.)

Figure 2.

PCA plot showing the distribution of the CRC tumor samples in two dimensional spaces into six subtypes. The upper and lower panels in the plot display the sample distribution using “citccmst discovery dataset” and our “input dataset,” respectively.

Figure 3.

Comparison of subtype proportion from our CRC (“internal,” red bar) dataset with that of the French (“citccmst,” green bar) dataset. (A color version of this figure is available in the online journal.)

Figure 4.

Survival plot showing the overall survival distribution of six predicted subtypes of CRC patients.

Figure 5.

Survival plot showing the overall survival distribution of low and high risk subtype groups. (A color version of this figure is available in the online journal.)

Figure 6.

Differential gene expression analysis of 57 genes in our CRC dataset. Red solid circles represent 22 out of 57 genes found differentially expressed in the CRC tumor dataset. (A color version of this figure is available in the online journal.)

Figure 7.

Ingenuity pathway analysis of 57 gene signature showing “cancer” as the most significant function associated with these genes. (A color version of this figure is available in the online journal.)

Figure 8.

Top scoring network containing 11 out of 57 genes indicating the association with cancer, hematological disease and immunological disease. (A color version of this figure is available in the online journal.)