Exploring the importance of cancer pathways by meta-analysis of differential protein expression networks in three different cancers

It is known that for most cancers with solid tumors the genes in the above mentioned “target pathways” display somatic mutations and change their protein products [1] . In all human tumors, PI3K is known to be as one of the most frequently targeted pathways. Mutation in PI3K pathway components contributes up to 30% of all human cancer and is known to be activated by RAS [14]. It is interesting to know that PI3K is a regulatory subunit, which binds to cell-surface receptors and to the RAS protein. Genes and proteins in PI3K and RAS have been investigated as therapeutic targets for many cancers [15, 16]. Here this purely quantitative analysis of the existing protein expression data of three different cancers also reveals the significant alteration of the proteins in PI3K and RAS pathways. Our findings are consistent with this and suggest that continued future efforts be made in this direction.

Since genes act in consort during a biological process, a network analysis is essential for a system-wide understanding. Thus, a study of differential network connectivity could yield interesting findings that are not possible from a differential expression analysis of individual proteins. In addition, pathway level information should be incorporated in a differential network analysis whenever possible. It can be seen from Fig. 3(d) that on the basis of analysis of differential connectivity of individual proteins, PI3K emerged as the most important pathway overall. However, this could be partly due to the fact that PI3K pathway has the largest number of proteins compared to the other “target pathways”. On the other hand, the test statistic based on differential connectivity of pathways is automatically normalized by the size of the pathway and on the basis of this comparison, RAS turned out to be the most significant overall as shown in Table 2.

In a recent study [17], multiple molecular profile data of LUAD for the CAMDA 2014 lung adenocarcinoma challenge data provided by ICGC is analyzed and it is noted that EGFR signaling pathway plays a significant role among the patients. Besides, it is known that EGFR activation initiates RAS signaling [18–20], and EGFR induces rapid increase in number of epithelial cells by activating a network of signaling elements, including members of the RAS and PI3K [20, 21]. Thus, it is common for lung cancer patients to have active EGFR mutation. Moreover, the RTK/RAS/RAF pathway, identified as the main route in causing adenocarcinoma, is shown to be activated in patients with “common” pathway mutations, e.g., in KRAS, BRAF, and EGFR [22]. Although the EGFR kinase inhibitor Gefitinib is an effective treatment for lung cancers with EGFR activating mutations, amplification of MET causes Gefitinib resistance by driving ERBB3 (HER3)–dependent activation of PI3K [23], a pathway thought to be specific to EGFR/ERBB family receptors. This fact, along with the fact that MET has been found in the module of the “progression” group but not in that of the “complete remission” group in PI3K signaling pathway of LUAD (see Fig. 2) in our current study, suggest that the patients under study might have been treated with Gefitinib, but the presence of MET in some of these patients (those in the “progression” group) may have led to the resistance to this drug. However, this is subject to verification as we don’t have the information regarding the treatment regime for any of these patients. This highlights the potential role of MET in lung cancer progression.

An interesting observation from Fig. 3 is that much lesser proportion of proteins in cell cycle pathway is differentially connected between the two groups of patients in KIRC compared to LUAD and HNSC. So underlying molecular mechanisms related to cell cycle pathway may be a little different in KIRC compared to LUAD and HNSC.

In a recent paper [24], patient level information such as mutation profiles is incorporated to identify protein-protein interaction (PPI) interfaces enriched in somatic mutations. It will be interesting to explore how to incorporate patient heterogeneity information into our approach.