Cancer is a multifactorial disease. In order to characterise cancer cells thoroughly, analysis of multi-omics data is necessary. Here, we are combining chromatin interaction profile and epigenome profile of oral cancer cells in order to study the drug response and metastasis in them. We are trying to answer the questions- Why do cells have different drug-response? What are all the genomic feature which decides a cancer cell's fate upon introduction of the different types of drugs? Can we discovery prominent target-sites for the drugs
A large part of the genome is non-coding in nature, especially the long non-coding RNA (lncRNA) genes, which have been annotated extensively because of the advances in the sequencing technologies, but their function is mostly unknown, some of them are associated with an array of functions- X chromosome inactivation, chromatin modification, transcription activation, nuclear transport. There has been several experimental as well as computational approach to unravel the role of the lncRNA genes which have met less success due to lack of uniformly shared feature across the genome. To negate this, we have utilized multi-omics data, which is relatively consistent across the genome, to build models to predict the ontology-based functions of the lncRNA genes.
We are trying to develop new analysis techniques for single expression and epigenome profiles and trying to extend application of single cell genomics beyond classical applications of finding heterogeneity and rare cells. We are analysing single cell genomic profiles with new perspectives so that it could help in development of novel diagnostic techniques and prediction of response to environment. One such application which can help doctors in finding matching cells for taking clinical decisions is the first single cell expression search engine (http://cellatlassearch.com/) made by us.
We published first comprehensive study on H2BK20ac (1 out 35 histone acetylations) and showed it's uniqueness in detecting cell state specific promoter activity and enhancers (Kumar et al., Genome Research). We are exploring its applications in diagnostics. With a grant on cell free DNA analysis, we are trying to find new markers for early cancer detection and developing new methods for reducing error in prediction of diseases using cfDNA.
We view the complexity in biological data-sets as an opportunity to develop generalized methods which can be useful for other fields also. Hence we are interested in developing generalized methods for bioinformatics projects involving signal processing, NP-complete, classification and feature-selection problems which could have wide application in other fields also. For example epigenome analysis has now become a signal processing problem. Similarly genome assembly involves a lot of steps which could be NP-complete. Analysis of single cell expression involves feature selection and classification. Currently we are developing computational methods for, analysing single cell data-sets, network inference and new applications of Graph signal processing.