Optimized prediction of extreme treatment outcomes in ovarian cancer

Abstract

Ovarian cancer is the fifth leading cause of death among female cancers. Front-line therapy for ovarian cancer is platinum-based chemotherapy. However, the response of patients is highly nonuniform. The TCGA database of serous ovarian carcinomas shows that ∼10% of patients respond poorly to platinum-based chemotherapy, with tumors relapsing in seven months or less. Another 10% or so enjoy disease-free survival of three years or more. The objective of the present research is to identify a small number of highly predictive biomarkers that can distinguish between the two extreme responders and then extrapolate to all patients. This is achieved using the lone star algorithm that is specifically developed for biological applications. Using this algorithm, we are able to identify biomarker panels of 25 genes (of 12,000 genes) that can be used to classify patients into one of the three groups: super respond-ers, medium responders, and nonresponders. We are also able to determine a discriminant function that can divide the entire patient population into two classes, such that one group has a clear survival advantage over the other. These biomarkers are developed using the TCGA Agilent platform data and cross-validated on the TCGA Affymetrix platform data, as well as entirely independent data from Tothill et al. The P-values on the training data are extremely small, sometimes below machine zero, while the P-values on cross-validation are well below the widely accepted threshold of 0.05.