One of the fundamental problems in population genetics and molecular evolution is to understand the drivers of genetic change in a population: which mutations affect the ability of an organism to survive, reproduce, and pass its genes to the next generation, while which mutations are mere "passengers" that do not affect this ability? This question is of great practical importance, for example, it can aid in the identification of mutations in viral genomes that enable viruses to evade immune response, in bacterial genomes that enable bacteria to develop drug resistance, and in mammalian cells that lead to cancer. In principle, the evolutionary history of a population contains information of the effects (deleterious, beneficial or neutral) of mutations occurring in the population...[ Read more ]

One of the fundamental problems in population genetics and molecular evolution is to understand the drivers of genetic change in a population: which mutations affect the ability of an organism to survive, reproduce, and pass its genes to the next generation, while which mutations are mere "passengers" that do not affect this ability? This question is of great practical importance, for example, it can aid in the identification of mutations in viral genomes that enable viruses to evade immune response, in bacterial genomes that enable bacteria to develop drug resistance, and in mammalian cells that lead to cancer. In principle, the evolutionary history of a population contains information of the effects (deleterious, beneficial or neutral) of mutations occurring in the population. However, extracting this information is challenging because the evolutionary dynamics depend not only on the cumulative effects of many mutations, but also on the competition between many individuals in a diverse population. The inference problem is further complicated by the confounding in uence of drift, recombination, and epistasis. Despite intense interest, until now it has been unclear how to systematically disentangle the effects of individual mutations from such complex evolutionary dynamics.

This work developed the marginal path likelihood (MPL) framework, a path integral approach, to estimate the effects of mutations from genetic time-series data which takes into account the complex evolutionary dynamics, thereby solving a longstanding problem in population genetics inference. A key feature of this framework is that it allows to derive an analytical expression for the estimate of effects of mutations. The resulting solution is computationally efficient and accurate. The work also included a detailed performance comparison of the MPL estimate with seven state-of-the-art methods on synthetic data using extensive simulations. The MPL estimate was then used to analyze a half-genome length data of HIV evolution. The last part of the work extended the MPL framework to also account for epistasis between pairs of mutations. The performance of the resulting estimate was tested on synthetic data as well as on HIV within-host evolution data.