Patterns of serotype switching and replacement in Pneumococcus
I am currently working on a project to identify non-random serotype switching patterns observed in Streptococcus pneumoniae. There are more than 90 serotypes observed in nature for this bacterium, several of which are invasive and cause diseases in both children and adults. Pneumococcal Conjugate Vaccine (PCV) target these invasive serotypes, which has led to increase in carriage of non-vaccine serotypes. This replacement is mostly driven by recombination which sometimes leads to switching of serotypes from vaccine-targeted to vaccine-escaping, while maintaining their invasiveness. I am using a database of more than 37000 clinical isolates and using analytic methods like Monte Carlo simulations and Market Basket Analysis to identify novel serotype switching patterns. To identify genes and pathways that might provide mechanistic insight into this phenomenon, our lab has collected whole genome sequences from different serotypes. To analyze this data, I have written NGS pipelines for variant calling and have parallelized the code to increase speed and efficiency.
Genetic and phenotypic predictors of sleep behavior in mice
My PhD research involved working on the international Knockout Mouse Phenotyping Program (KOMP2). The aim of the project is to develop mice with null mutations in single genes and study them on a broad-based phenotyping pipeline. The KOMP2 data consists of measurements in several sleep variables from more than 300 knockout strains from more than 7000 mice including controls. To study the effect of genotype on individual variables, I performed ANOVA with Dunnet’s post hoc analysis. And to study the effect of genotype on overall sleep, Principal Component Analysis (PCA) and clustering was used to identify candidate genotypes. In the next step, I gathered data for all phenotypes from KOMP2 database through APIs. The dataset consisted of 34 million data points. I performed Bayesian imputation for missing data and implemented machine learning algorithms like decision trees and random forest to identify sleep-disorder related phenotype predictors.
Meditation and Performance
Previous research from our lab had shown that meditation improves psychomotor performance. This study was conducted in both novice and experienced meditators. In the next step, we focused the study on novice meditators with 20 minute bouts of meditation done at different times of the day. We also did EEG recordings to identify correlates of brain activity and performance. I was able to develop the project from scratch through a successful grant which allowed us to acquire a portable EEG system for recordings during meditation. I wrote the IRB proposal, and managed every stage of the project while mentoring two undergraduates, and co-instructed for a course for 15 students who worked on the project as well.