Current research in statistical genetics
Discovery of trans-eQTLs
Genetic variants which regulate distant target genes are called trans-acting expression quantitative trait loci – in short, trans-eQTLs.
Many genetic variants are believed to mediate disease risk via the trans-eQTLs.
Discovering trans-eQTLs and understanding their mechanism is crucial for the future of personalized or precision medicine.
Since 2018, I have been working on developing new statistical methods, with a focus on discovering trans-eQTLs.
Bayesian regression requires computing the often intractable posterior probability.
Given any prior and likelihood, the quasi-Laplace approximation implicitly assumes the posterior as a Gaussian
whose mode and variance is close to that of the exact posterior.
If the prior is Gaussian, it reduces to the well-known Laplace approximation.
Logistic regression with quasi-Laplace approximation, as implemented in B-LORE,
outperforms existing state-of-the-art models for finemapping case-control GWAS.
Previous research in statistical physics
Diffusion in a rough potential.
Rugged energy landscapes find wide applications in diverse fields ranging from astrophysics to protein folding.
We found a general expression for the effective diffusion coefficient of a Brownian particle on a random energy surface.
This general expression reduces to the famous equation of Zwanzig in the limit of infinite spatial correlation!
Orientational order leads to long-range hydrophobicity.
The long range attractive force between two hydrophobic surfaces immersed in water is observed to decrease exponentially with their separation – this distance-dependence of effective force is known as the hydrophobic force law. Although the phenomenon is conventionally explained by density-dependent theories, we identified orientation, rather than density, as the relevant order parameter using the 2D Mercedes Benz model of water. The range of density variation was noticeably shorter than that of orientational heterogeneity.
Percolation transition of amphiphilic solutes in aqueous binary mixtures.
Aqueous binary mixtures of many amphiphilic solutes such as methanol, dimethyl sulfoxide (DMSO), ethanol, dioxane, phenol, glycerol, etc. show dramatic, often exotic, anomalies in thermodynamic and dynamic properties. In a series of work during my PhD, we found a percolation transition of the solutes, leading to the formation of a spanning cluster of self-aggregates in the system after a critical percolation concentration. Surprisingly, the onset of this percolation occurs at a concentration at which the thermodynamic and dynamic anomalies are found experimentally.