Diverse molecular data from a large group of patients with multiple types of tumors can be combined with clinical variables to help predict patients’ overall survival durations, a recent study found. The study also showed that a large-scale analysis of multiple tumor types can detect important genetic alterations that would likely be missed in an analysis of a single tumor type.
In the multi-institutional study, researchers analyzed several types of molecular data (somatic copy-number alteration; DNA methylation; and mRNA, microRNA, and protein expression) from 953 samples of four cancer types (clear cell renal cell carcinoma, glioblastoma multiforme, ovarian serous adenocarcinoma, and squamous cell carcinoma of the lung). Also studied were clinical variables such as patient age and tumor stage from the same set of patients. The molecular and clinical data came from The Cancer Genome Atlas.
Statistical models developed by Han Liang, Ph.D., an assistant professor in the Department of Bioinformatics and Computational Biology at The University of Texas MD Anderson Cancer Center, and his colleagues showed that the combination of molecular data and clinical factors predicted overall survival more accurately than did clinical factors alone in three of the four cancer types studied (renal, ovarian, and lung cancer).
“In contrast to previous studies driven by a single cancer or data type, we could evaluate patient survival prediction from different molecular data and describe the potential prognostic and/or therapeutic relevance across multiple cancers,” Dr. Liang said.
An additional analysis of 12 cancer types identified 10,281 somatic alterations in clinically relevant genes. Dr. Liang said that in an analysis of a single tumor type, many of these alterations would not have been revealed at a frequency that would justify further investigation.
“By analyzing data from multiple cancer types, we were able to systematically evaluate prognostic models and more thoroughly identify gene alterations that led to tumor formation,” Dr. Liang said. “This would have not been obtained by looking at tumor data from just one cancer type.”
Dr. Liang added that future studies of independent patient cohorts will be needed to determine how large-scale molecular profiling data might be used with clinical variables to stratify cancer patients into various risk groups to help guide surveillance and treatment strategies.
The study’s report was published in June in the online version of Nature Biotechnology.
OncoLog, September 2014, Volume 59, Issue 9