Gene Mutation Screening Upstream Of Biomarkers Has The Potential To Identify And Fix Compromized Conclusions

The purpose of this study is to assess potential problems of single-nucleotide variant (SNV) biomarker testing for cancer treatment decision support. Specifically, we investigate whether mutations upstream of predictive biomarkers have the potential to invalidate the predictive conclusions drawn on the basis of just testing for the biomarker mutation itself and neglecting the context. We investigate 25 cases of endometrial cancer (EC) with whole exome sequencing data (from TCGA), and 23 cases of breast cancer (BC) sequenced for a panel of u003e600 cancer-related genes. Each case is investigated for the presence of any of 1825 SNVs that have been curated from the medical literature for potentially being predictive of treatment effect in at least one cancer type. For each of these SNVs, we look for other mutations that may compromize the biomarker by destroying the corresponding protein, specifically upstream frame-shift (fs) insertions or deletions (indels), splice-site disruptions (SSDs), and premature stop codons (PSCs). We count two types of dubious constellations. First, biomarker mutations found present along with a dramatic upstream mutation in the same gene are potentially false positive (FP), because altering a destroyed protein may be without impact. Second, where a biomarker mutation is absent, an equivalent effect may be caused by a deleterious upstream mutation, such that the situation amounts to a potential false negative (FN). Looking for misleading presence of biomarkers (FPs), we find potentially disruptive mutations upstream of known biomarkers in 5 of our 48 cases. Closer investigation suggests, however, that none of them are real FPs. The most important reason is that, for biomarkers that inactive the molecular function of the protein, any further damage to the protein would rather strengthen the effect. In contrast, potentially misleading absence of any inactivating biomarker in a gene that is disrupted by a different mutation is frequent in our cohort. Starting from a list of 716 human tumor suppressor genes taken from TSGene (Vanderbilt), we find that 23 have at least one associated biomarker SNV that is absent and at the same time have a deleterious upstream mutation in at least one patient in our cohort. For a more detailed investigation we focus on P53. In the EC cohort, we have 1 case with a known P53 biomarker, and among the 24 others 1 case with upstream disruptions (both fs-indel and SSD). In the BC set, there are 2 regular P53 biomarker calls, and 4 additional cases without biomarker mutation but with upstream disruptions (all fs-indels, 1 also with PSC). Reassuringly, the conclusions based on biomarker mutations found to be present seem to be reliable, at least in the investigated cohorts. However we do find disruptions upstream of called biomarker mutations in ~10% of the investigated patients, which demonstrates that there is a real danger of false positive findings. Just looking at P53, we find that considering disruptions in addition to the described biomarkers (3 cases), such that they formally differ from the biomarkers but may have similar consequences (5 cases), almost triples the number of hits. In contrast to hot-spot sequencing and SNP arrays, NGS-based whole gene sequencing allows for holistic assessment of biomarker genes. It consequently may allow more precise cancer diagnostics and may benefit treatment decision. This study may be expanded in several ways, including taking into account variant allele frequencies and investigating larger cohorts. In particular we plan to include metastatic cancer and cases of acquired resistance, which may have increased incidence of mutations that interfere with biomarkers. Citation Format: Sonia Vivas, Francesca Diella, Alexander Zien. Gene mutation screening upstream of biomarkers has the potential to identify and fix compromized conclusions. [abstract]. In: Proceedings of the AACR Special Conference on Translation of the Cancer Genome; Feb 7-9, 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 1):Abstract nr A1-60.