Evolution and Immunology of pathogens

Pathogens evolve rapidly to circumvent drug treatments and immune surveillance, which dramatically impacts public health. Research and treatment are complicated by high genetic diversity of some viruses within and across infected individuals, as well as their complex evolutionary mechanisms, including selection, random genetic drift, and temporal variation in a host environment. Moreover, many pathogens have a large number of linked sites approximately 102 -103 for HIV and hepatitis C virus (HCV) that evolve simultaneously and inter-dependently through two different effects, "epistasis" due to interaction between proteins and signaling network, and co-inheritance linkage ("clonal interference").

Our research is focused on developing mathematical tools that predict evolution of pathogens with strong linkage effects, including analytic and computational methods and estimators of evolutionary parameters from sequence data.


The last decade has seen explosive progress in mathematical modeling of microbial populations and high-fidelity sequencing. Taking advantage of these developments, my team will address evolution of microbes (yeast, bacteria) and viruses (HIV, influenza, polio, CHIKV, Dengue, HCV). Launching from my previous mathematical and applied studies, we are applying existing methods and models to study the viral evolution under time-dependent conditions, develop new mathematical techniques and improve existing phylogenetic tools, and identify some key factors of HIV pathogenesis. Our multi-disciplinary team fuses the recent mathematical discoveries with multiple-scale modeling and software tools. We are especially interested in the evolutionary effects of epistasis, recombination, and the theory of phylogenetic relationships in the presence of selection and the other factors. The project is designed to create significant clinical impact by fostering research into novel classes of drugs to control viral adaptation rate and achieve viral containment. Our software will facilitate personalized medicine and vaccine design against the pathogens escaping treatment and immune responses. The results are published and diffused in higher education and public presentations.

High-fidelity detection method of epistasis

Pedruzzi and Rouzine 2020

The computer model of asexual evolution includes the factors of random mutation, selection, epistasis, and random genetic drift. Pairwise haplotype frequencies f_ij are averaged over an ensemble of independent simulation runs (populations). The strength of interactions, UFE_ij, is calculated. The indirect links and the residual stochastic LD are excluded by using triple-site haplotype frequencies, UFE_ij0. B.Pre-set epistatic network for 50 sites.Real epistatic linksare shown by green lines. The resulting indirect links are red lines. Some examples of stochastic linkage bonds are shown by blue lines. C-D. The network of strong (UFE > 0.5) candidate epistatic interactions predicted (C) from a single population and (D) after averaging over 200 populations. E. Scatter plot of 3-locus haplotype min(UFE_ij0) shown against UFE_ij for the pairs identified in (D). Dashed sector (green): Predicted direct interactions. F. Predicted network accurately recapitulates the pre-set epistatic network. Parameters: initial allele frequency 0.45, mutation rate per genome U=0.07, fixed selection coefficient s=0.1, epistatic strength E=0.75.

Main achievements

(2020) Book "Mathematical models of evolution" published by De Gruyter reviews basic one-locus theory, asexual multi-locus evolution, and evolution with recombination. The factors included are mutation, natural selection, and random genetic drift.

(2020) The omnipresent exponential spectrum of beneficial mutations in fitness effect observed in viruses and microbes is explained. The intrinsic spectrum measured by directed mutagenesis and the observed spectrum in evolution experiments are shown to differ.

(2020) It is shown that averaging over 100 independent populations is necessary but not sufficient to detect epistasis, due to a high residual number of false-positive interactions. A triple-way haplotype method is developed to compensate for this noise and detect the true intetactions.

(2019) Detection of epistasis from a DNA sample of a single evolving population is shown to be impossible regardless of the method used, which explains the high false-positive rate observed experimentally

(2018) A direct measure of epistatic interaction in terms of Darwinian fitness from the haplotype frequencies of a genomic site pair is developed

(2018) A model describing the antigenic evolution of a virus in a host population with immune memory is analyzed and compared to data for influenza A

(2016) The theoretical prediction of the existence of an adaptation optimum in the mutation rate is confirmed by experiments on polio virus in mice (collaboration with UCSF)

(2015) A model for the Trojan horse effect of virus latency in HIV transmission is analyzed and compared to data from patients

(2013, 2016) Virus escape from the treatment with a defective interference particle is investigated

(2012) The traveling wave theory is generalized for arbitrary distribution of mutation effect on fitness (collaboration with UCSB, Harvard, and U. Goettingen)

(2005-2010) Recombination is incorporated into the traveling wave theory

(2003) A traveling wave theory is developed to predict the adaptation of asexual populations

(2003) The nature of parasitemia oscillations in malaria is investigated using an age-structured model with inter-cell communication

(1999, 2011) Evolutionary parameters of HIV are estimated (effective population size, average selection coefficient, recombination rate)

(1999) Rapid evolution and high diversity of HIV is explained from the emergence of mutations compensating primary mutations conferring escape from the immune recognition

(1999-2001) First introduction of stochastic models of evolution in the presence of selection into virology