Pal Research Group:

Google Scholar

Research Areas

Machine Learning 

Recently, we have explored various aspects of regression modeling such as sequential feature selection, optimal stacking, inconsistencies in databases, functional regression, Copula based Multivariate Random Forest design, Heterogeneity Aware Predictive Modeling and diffusion modeling

REFINED (REpresentation of Features as Images with NEighborhood Dependencies): A novel feature representation for Convolutional Neural Networks in press with Nature Communications arxiv 

Drug Sensitivity Prediction 

In recent few years, we have been exploring the issue of drug sensitivity prediction from genomic and functional characterizations. We were a top performer in NCI-DREAM drug sensitivity prediction challenge from genomic characterizations and the methodology was published in PLOS One and Nature Biotech. In collaboration with Dr. Charles Keller, we have also designed an alternative framework based on integrated functional and genomic characterizations that has excellent performance in terms of predicting untested drug sensitivities and synergistic drug combinations. The studies reported in multiple publications including one in Nature Medicine demonstrate the potential of our methodology to select optimal drug combinations for cancer patients. Recently, in collaboration with Dr. Souparno Ghosh, we have also explored various other aspects of predicting drug sensitivity such as sequential feature selection, stacking and inconsistencies in databases. 

 

Genetic Regulatory Network Modeling and Control

The article on IEEE Signal Processing Magazine provides a concise overview of our research on robustness in genetic regulatory network modeling and control (pdf).  

 

  • Studying the effect of coarse-scale modeling

        

        Representative Publications:  IEEE TSP, Vol. 58, Pg. 3341 - 3351, No.6, 2010; IEEE SPM  Vol. 29, No. 1, Pg. 66-76, 2012

  

 Funding