I was initially trained as a theoretical physicist with a PhD from India. Before coming into computational biology-oriented, I worked on various topics in plasma physics, condensed matter physics and nano-materials. Inspired by the applied research in health sciences, I became interested in computational biology and bioinformatics. In 2009, I joined Bioinformatics Division at Center for Cancer Research and Cell Biology, Queen’s University Belfast as a post-doc. I developed a combinatorial optimization algorithms for microRNA target ranking and landscape analysis in RNA folding simulations. Collaborating with Applied Mathematics groups at Queen’s Belfast, I also developed stochastic models for avascular tumour growth. In 2011, I joined Jakob Skou Pedersen’s group at Department of Molecular Medicine (MOMA), Aarhus University as a member of COAT.
My research is on transcriptome-wide RNA structure mapping. The initial goal was to develop a flexible probabilistic model (SCFG-based) for RNA secondary structure prediction that integrates sequence data and NGS-based probing data. A computational method is now in place, which is capable of predicting RNA secondary structures with constraints from SHAPE chemical probing data. This model can be extended to integrate various other types of structure probing as well. As an example, the following ROC curve shows the performance of our models in predicting E. Coli 16S and 23S rRNAs with constraints from SHAPE data:
ROC curve for E. coli 16S and 23S using our “pair model” (black curve) and “stacking model” (red curve); The “pair model” does not account for base-stack interactions. The F-value identifies the point on the ROC curve where the overall performance is maximized. The performance of both these models increases with SHAPE data (F-values about 19% higher for 16S rRNA and about 9% higher for 23S rRNA).