ParSel: Parallel Selection of Micro-RNAs for Survival Classification in Cancers

Mol Inform. 2017 Jul;36(7). doi: 10.1002/minf.201600141. Epub 2017 Feb 13.


It is known that tumor micro-RNAs (miRNA) can define patient survival and treatment response. We present a framework to identify miRNAs which are predictive of cancer survival. The framework attempts to rank the miRNAs by exploring their collaborative role in gene regulation. Our approach tests a significantly large number of combinatorial cases leveraging parallel computation. We carefully avoided parametric assumptions involved in evaluations of miRNA expressions but used rigorous statistical computation to assign an importance score to a miRNA. Experimental results on three cancer types namely, KIRC, OV and GBM verify that the top ranked miRNAs obtained using the proposed framework produce better classification accuracy as compared to some best practice variable selection methods. Some of these top ranked miRNA are also known to be associated with related diseases.

Keywords: cancer survival; classification; feature selection; miRNA prediction.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Algorithms
  • Biomarkers, Tumor
  • Computational Biology / methods*
  • Databases, Genetic
  • Gene Expression Profiling
  • Gene Expression Regulation, Neoplastic
  • Gene Regulatory Networks
  • Humans
  • MicroRNAs / genetics*
  • Neoplasms / genetics*
  • Neoplasms / mortality*
  • Software*
  • Workflow


  • Biomarkers, Tumor
  • MicroRNAs