Applying Computational Solutions for solving problems in Mammalian Gene Family Evolution and Single Cell Gene Expression Analysis

Using computational tools to solve various biological problems has become common practice over the last decade. This has been primarily fueled by exponentially growing high throughput biological data and relevant computational biology resources to support it [1]. The work presented in this thesis showcases application of computational techniques to answer critical biological questions pertaining to gene family evolution and single cell gene expression analysis. Dr. Susumu Ohno was a pioneer to propose the significance of gene duplication in driving gene family evolution. Gene duplication has been shown to expand the repertoire of several gene families involved in multitude of important biological functions. Thus, understanding the forces that lead to retention and loss of gene duplicates have been subject to close scrutiny over the past decade. One of the works presented in this thesis provides a strong argument for a novel gene duplication model that seeks to explain retention of previously unexplained gene duplicates. The first part of the work involved an in-depth study of evolutionary history of mammalian ribosomal protein gene family (RPG) evolution. This study confirmed our prior preliminary finding that there are thousands of intact duplicates whose fate could not be explained by existing gene duplication models. This led us to frame a novel gene duplication model that explains the retention of these gene duplicates. We also make a strong case for this model by employing rigorous in-silico and in-vitro tests to demonstrate its feasibility. We have achieved a rare feat by employing the above-mentioned two orthogonal but necessary tests that are lacking in other gene duplication models. Our investment in studying RPG family evolution enabled us to frame a novel single cell RNA-Seq (scRNA-Seq) QC pipeline. We hypothesized that the biological constraints under which RPGs function could serve as a robust biological indicator for cell health at single cell resolution. We formulated an outlier based QC model consisting of three features that could be extracted from RPG transcriptional signatures in any scRNA-Seq dataset. We show stable performance of the model across various datasets along with comparison with other QC features widely used in existing approaches. This QC model is designed to be easily implemented and applied to any scRNA-Seq study irrespective of experimental approach.