Gene Cluster Expression Index (GCEI) and Potential Indications for Targeted Therapy and Immunotherapy

Lung cancer recurrence risk was demonstrated to be related to driver gene and immunotherapy target gene cluster expression abnormality. Nine clusters seeded with driver genes ALK, BRAF, EGFR, MET, NTRK, RAS, RET, ROS1, TP53 and two immunotherapy target genes PDCD1 and CTLA4 were investigated respectively to predict lung cancer recurrence. The cluster of a seed was pre-selected to include fusion partner genes in the case of gene fusion, ligands, its pseudogenes, upstream and downstream co-expressors or inhibiting genes, effectors directly related to important pathways, etc. For each cluster, a gene cluster expression index (GCEI) was defined in two steps: Firstly, apply the univariate ROC of using each members expression vector to predict recurrences to label a patient sample as either normal or abnormal; Secondly, apply the percentage of abnormal genes in the cluster to predict recurrences to derive an optimal threshold so that a cluster member voting strategy can be achieved and a sample is labeled as abnormal (with respect to the cluster expression profile) if the the percentage of abnormal genes for the sample is greater than or equal to the threshold and as normal vice versa. Combinatory GCEI was developed as a binary string concatenating the individual GCEI corresponding to the individual cluster in an ordered list of driver or other important gene seeds. It showed that the recurrence risk of the abnormal group is typically 50% to 200% higher than the normal counterpart. Finally it was proposed and discussed to expand targeted therapy and immunotherapy to the abnormal group defined by GCEI. BackgroundMolecular profiling such as DNA-based mutation panels and proteiomics have been demonstrated great success in oncology for personalized medicine. Transcriptome profiling has emerged to be another promising opportunity as complement and expansions to the DNA-based approach and as new tools to further advance clinical oncology. MethodsLung cancer gene expression GEO data sets were downloaded, normalized, combined and analyzed. A novel approach was presented to analyze expression abnormality of important gene clusters with seeds including drivers such as ALK, BRAF, EGFR, MET, NTRK, RAS, RET, ROS1, TP53 or immunotherapy target PDCD1 and CTLA4, etc. A cluster was pre-specified for each seed and included the fusion partners in the case of translocation, ligands, activators, inhibitors, effectors, co-stimulators in the important pathways, etc. Each cluster member was labeled as normal or abnormal (up or down) with the univariate ROC by using its expression to predict recurrences. Cluster level labeling of expression state (normal or abnormal) was via a dynamic voting strategy, of which the voting threshold was set as the optimal cutoff on the ROC associated with the univariate model of using the percentage of the abnormal members to predict recurrences. Given an ordered list of important genes, a binary string of the same length was encoded by assigning 0 for normal and 1 for abnormal representing the cluster expression state of the corresponding position, called gene cluster expression index (GCEI) signature. Finally lung cancer recurrences were assessed and compared based on GCEI states and the combinations. ResultsThe recurrence risks of single gene normal group (GCEI = 0) vs abnormal group (GCEI = 1) were as follows, ALK: 17% vs. 55% for all stages, 13% vs. 42% for Stage I, 36% vs. 67% for Stage II-IV; BRAF: 23% vs. 49% for all stages, 15% vs. 36% for Stage I, 54% vs. 59% for Stage II-IV; EGFR: 25% vs. 47% for all stages, 17% vs. 33% for Stage I, 54% vs. 59% for Stage II-IV; MET: 25% vs. 44% for all stages, 17% vs. 29% for Stage I, 51% vs. 60% for Stage II-IV; NTRK: 19% vs. 52% for all stages, 13% vs. 40% for Stage I, 44% vs. 63% for Stage II-IV; RAS: 24% vs. 51% for all stages, 16% vs. 35% for Stage I, 47% vs. 65% for Stage II-IV; RET: 19% vs. 50% for all stages, 14% vs. 35% for Stage I, 40% vs. 65% for Stage II-IV; ROS1: 23% vs. 48% for all stages, 17% vs. 32% for Stage I, 45% vs. 64% for Stage II-IV; TP53: 23% vs. 50% for all stages, 15% vs. 38% for Stage I, 49% vs. 64% for Stage II-IV; and for the immunotherapy target gene: CTLA4: 26% vs. 49% for all stages, 14% vs. 38% for Stage I, 53% vs. 62% for Stage II-IV; PDCD1: 28% vs. 48% for all stages, 16% vs. 37% for Stage I, 54% vs. 61% for Stage II-IV. In addition, taking 9-driver gene GCEI and summarizing number of 1, the count of abnormal driver genes, N, and then comparing the population of N [≤] 5 vs. N > 5, the recurrence risks were: 19% vs. 59% for all stages, 13% vs. 49% for Stage I, 41% vs. 66% for Stage II-IV. Hence most of the cases the recurrence risk is 1.5 to 3 times higher for patient group with abnormally expressed gene clusters than normally expressed. DiscussionPrecision medicine based on RNA expression analysis is discussed and it is conjectured to apply targeted therapy or immunotherapy to lung cancers based on the related gene expression status as determined by the cluster member voting strategy. This can serve as an extension and complement to the current DNA-based tests, especially for a majority of patients who have been tested negative based on the conventional tests and have possibly missed the potential treatment benefit.