Abstract 4267: Patterns of DNA copy-number alterations revealed by the GSVD and tensor GSVD encode for cell transformation and predict survival and response to platinum in adenocarcinomas

Abstract

Abstract A quarter or more of lung, uterine, and ovarian adenocarcinoma (LUAD, USEC, and OV) tumors are resistant to platinum, the first-line systemic treatment for decades. Only recently, however, patterns of copy-number alterations (CNAs) were discovered that are predictive of OV survival, both in general and in response to platinum, and only by using a comparative spectral decomposition, the tensor generalized singular value decomposition (GSVD). We defined the tensor GSVD to extend the GSVD from two matrices to two tensors. Like the GSVD, the tensor GSVD can simultaneously identify the similar and dissimilar between two datasets recording different aspects of a single phenomenon (1, 2). In a comparison of Agilent microarray platform-matched profiles of patient-matched normal and primary, mostly high-grade OV DNA, the tensor GSVD revealed the tumor-exclusive and platform-consistent patterns across the chromosome arms 7p and Xq and the combination of 6p+12p. No other diagnostic can distinguish a shorter from a longer survival in response to platinum throughout the course of an adenocarcinoma. Here we show that the 6p+12p genotype predicts a survival phenotype in LUAD and USEC in addition to OV. We use the GSVD to compare whole-genome sequencing (WGS) and Affymetrix microarray profiles of patient-matched normal and primary LUAD and primary, mostly high-grade USEC and OV DNA. The GSVD reveals tumor-exclusive patterns of CNAs across the combination of the chromosome arms 6p+12p. We find that, first, like the Agilent OV pattern, the WGS LUAD and Affymetrix LUAD, USEC, and OV patterns identify a biologically consistent shorter survival phenotype, of approximately one and a half years median survival time in LUAD and USEC and three years in OV. Like the Agilent pattern, the WGS and Affymetrix patterns describe a biologically consistent genotype, where loss of the p21-encoding CDKN1A and p38-encoding MAPK14 on 6p, and gain of KRAS on 12p, together, but not separately, encode for human cell transformation. Second, by classifying the WGS and Affymetrix profiles, the Agilent pattern proves to be a technology-independent predictor of a LUAD, USEC, or OV patient's survival, both in general and in response to platinum. The pattern is independent of the tumor's stage, the best indicator of these adenocarcinomas at the time of initial diagnosis. Third, the pattern is a predictor of OV survival and response to platinum beyond the time of initial diagnosis and throughout the disease, even in patients experiencing remission after treatment of the primary tumor, and independent of the time to recurrence or progression. We conclude that comparative spectral decompositions, such as the GSVD and tensor GSVD, underlie a mathematically universal description of the relations between a tumor's genome and a patient's survival and response to treatment, which other methods miss.