There are several ongoing projects analyzing largescale datasets including - gene expression, metabolomics, single-cell RNA seq, proteomics, serum cytokine levels and known information about signaling and metabolic pathways. Recently, we are also working with adaptive immune receptor repertoire sequencing. A range of statistical and informatics tools are used for the analysis. For example - canonical correlation analysis, multiple correspondence analysis, structural equation modeling, sparse partial least square regression, etc.
Specific methodology developed in the lab is to integrate knowledge with the large-scale data in order to study human response. Networks are most intuitive way to depict information. Networks also known as graphs can be used to integrate knowledge with large-scale datasets. We have been studying molecular networks responsive to viral infections and vaccinations (e.g. Katanic et. al. Journal of Immunological Methods 2017). In recent collaborations with Department of Defense we have used integrative network analysis to link clinical outcomes measured by ICD9 (now ICD10) codes with molecular measurements (Thakar et. al. JOEM 2019). When network topologies cannot be inferred/ available context specific responses to infections, vaccinations or environmental exposures can be curated from data available in public domain. We have used Fuzzy-C-Means clustering (Khan et. al. BMC Bioinformatics 2017) along with other statistical and clustering approaches to identify context specific molecular signatures. One of the focus is also improving transcription factor target sets to define transcription factor activities (Mariani et. al. Journal of Infectious Disease 2017, Thakar et. al. BMC Immunology 2015). This curated information can be used with available statistical methods for pathway analysis. We have also developed a series of methods for pathway analysis (Zhang et. al. Bioinformatics 2017). The ongoing projects involve development of supervised methods to assemble networks informed by existing knowledge.
Modeling trajectories of host response: Networks/ graphs allow development of advanced computational approaches to investigate molecular interactions and their trajectories. Discrete-state models can be developed to investigate signal flow and integration through the molecular networks. These enhanced networks define integration of signals using logic-rules and can be simulated for different starting conditions which can be mapped to demographic variables such as age, gender. Taking into account specific starting conditions, the methods are developed to study trajectories mapping to different disease states. Details of this approach can be found in many of our publications (e.g. Thakar et. al. PLoS Comp Bio 2007, 2012, Walsh et. al J Immunology 2011, Anderson et. al. Computational and mathematical methods in medicine 2016, Palli et. al. PLoS Comp Bio 2019). Ongoing projects involve development of methodology for different omics datasets.