Discovery of ultra-low-signal variants to study human evolution and epilepsy

Abstract

Next-generation sequencing (NGS) has revolutionized research on human evolution and brain-related diseases. With DNA-sequencing data available from humans at different time periods and affected tissues in the disease of interest, researchers have been able to conduct more quantitative analyses to answer questions such as which genetic variants likely contribute to higher functionality of the human brain and which genetic variants occurring in which parts of the brain potentially cause epilepsy. However, challenges in variant detection and interpretation remain and require significant efforts to develop novel computational methods accompanied by experimental techniques. Highly degraded short DNA templates and non-even sequencing depth lead to difficulty in detection of transposable elements (TEs) from ancient humans, while somatic mutations, i.e., non-hereditary mutations that arise during human development, can be present at very low variant allele fractions and elude detection. In this thesis, I reliably detect underappreciated genetic variants due to their low signal-to-noise ratio from NGS data and reveal how these variants can contribute to the genetic basis of human evolution and different types of epilepsy. In Chapter 1, I show that our new machine learning method can reliably detect TEs in highly degraded and unevenly sequenced ancient human DNA, which provides some clue of how they can affect human evolution. In Chapter 2, I describe our efforts to capture weak signals of somatic mutations from hippocampus and present strong evidence that these mutations in cancer-related genes can cause temporal lobe epilepsy. In Chapter 3, I report our success in detecting known and novel somatic mutations from trace DNA adhering to single seizure-monitoring electrodes from pediatric epilepsy patient brains. These results have paved the way for alternative genetic explanations of human evolution, revealed the surprising crosstalk between cancer signaling pathways and non-tumor associated epilepsy, and advanced early and minimally invasive molecular diagnosis of epilepsy.