Biology subjects

Voznica, J.

Publications and source records attributed to Voznica, J..

9 recordsLinked to original sources

bioRxiv2025

Graph Neural Networks for Likelihood-Free Inference in Diversification Models

A common approach to infer the processes that gave rise to past speciation and extinction rates across taxa, space and time is to formulate hypotheses in the form of probabilistic diversification models and estimate their parameters from extant phylogenies using Maximum Likelihood or Bayesian inference. A drawback of this approach is that likelihoods can easily become computationally intractable, limiting our ability to extend current diversification models with new hypothesized mechanisms. Neural networks have been proposed as a likelihood-free alternative for parameter inference of stochastic models, but so far there is little experience in using this method for diversification models, and the quality of the results is likely to depend on finding the right network architecture and data representation. As phylogenies are essentially graphs, graph neural networks (GNNs) appear to be the most natural architecture but previous results on their performance are conflicting, with some studies reporting poor accuracy of GNNs in practice. Here, we show that this underperformance was likely caused by optimization issues and inappropriate pooling operations that flatten the information along the phylogeny and make it harder to extract relevant information about the diversification parameters. When equipped with PhyloPool, a new time-informed pooling procedure, GNNs show similar or better performance compared to all other architectures and data representations (including Maximum Likelihood Estimation) that we tested for two common diversification models, the Constant Rate Birth-Death and the Binary State Speciation and Extinction. We conclude that GNNs could serve as a generic tool for estimating diversification parameters of complex diversification models with intractable likelihoods.

evolutionary biology↗

bioRxiv2023

A Comparison of Deep Learning Architectures for Inferring Parameters of Diversification Models from Extant Phylogenies

AO_SCPLOWBSTRACTC_SCPLOWTo infer the processes that gave rise to past speciation and extinction rates across taxa, space and time, we often formulate hypotheses in the form of stochastic diversification models and estimate their parameters from extant phylogenies using Maximum Likelihood or Bayesian inference. Unfortunately, however, likelihoods can easily become intractable, limiting our ability to consider more complicated diversification processes. Recently, it has been proposed that deep learning (DL) could be used in this case as a likelihood-free inference technique. Here, we explore this idea in more detail, with a particular focus on understanding the ideal network architecture and data representation for using DL in phylogenetic inference. We evaluate the performance of different neural network architectures (DNN, CNN, RNN, GNN) and phylogeny representations (summary statistics, Lineage Through Time or LTT, phylogeny encoding and phylogeny graph) for inferring rates of the Constant Rate Birth-Death (CRBD) and the Binary State Speciation and Extinction (BISSE) models. We find that deep learning methods can reach similar or even higher accuracy than Maximum Likelihood Estimation, provided that network architectures and phylogeny representations are appropriately tuned to the respective model. For example, for the CRBD model we find that CNNs and RNNs fed with LTTs outperform other combinations of network architecture and phylogeny representation, presumably because the LTT is a sufficient and therefore less redundant statistic for homogenous BD models. For the more complex BiSSE model, however, it was necessary to feed the network with both topology and tip states information to reach acceptable performance. Overall, our results suggest that deep learning provides a promising alternative for phylogenetic inference, but that data representation and architecture have strong effects on the inferential performance.

ecology↗

bioRxiv2022

Deep Learning from Phylogenies for Diversification Analyses

Birth-death models are widely used in combination with species phylogenies to study past diversification dynamics. Current inference approaches typically rely on likelihood-based methods. These methods are not generalizable, as a new likelihood formula must be established each time a new model is proposed; for some models such formula is not even tractable. Deep learning can bring solutions in such situations, as deep neural networks can be trained to learn the relation between simulations and parameter values as a regression problem. In this paper, we adapt a recently developed deep learning method from pathogen phylodynamics to the case of diversification inference, and we extend its applicability to the case of the inference of state-dependent diversification models from phylogenies associated with trait data. We demonstrate the accuracy and time efficiency of the approach for the time constant homogeneous birth-death model and the Binary-State Speciation and Extinction model. Finally, we illustrate the use of the proposed inference machinery by reanalyzing a phylogeny of primates and their associated ecological role as seed dispersers. Deep learning inference provides at least the same accuracy as likelihood-based inference while being faster by several orders of magnitude, offering a promising new inference approach for deployment of future models in the field.

evolutionary biology↗

bioRxiv2021

Deep learning from phylogenies to uncover the transmission dynamics of epidemics

Widely applicable, accurate and fast inference methods in phylodynamics are needed to fully profit from the richness of genetic data in uncovering the dynamics of epidemics. Standard methods, including maximum-likelihood and Bayesian approaches, generally rely on complex mathematical formulae and approximations, and do not scale with dataset size. We develop a likelihood-free, simulation-based approach, which combines deep learning with (1) a large set of summary statistics measured on phylogenies or (2) a complete and compact representation of trees, which avoids potential limitations of summary statistics and applies to any phylodynamics model. Our method enables both model selection and estimation of epidemiological parameters from very large phylogenies. We demonstrate its speed and accuracy on simulated data, where it performs better than the state-of-the-art methods. To illustrate its applicability, we assess the dynamics induced by superspreading individuals in an HIV dataset of men-having-sex-with-men in Zurich. Our tool PhyloDeep is available on github.com/evolbioinfo/phylodeep.

bioinformatics↗

bioRxiv2021

Cuban history of CRF19 recombinant subtype of HIV-1

CRF19 is a recombinant form of HIV-1 subtypes D, A1 and G, which was first sampled in Cuba in 1999, but was already present there in 1980s. CRF19 was reported almost uniquely in Cuba, where it accounts for ~25% of new HIV-positive patients and causes rapid progression to AIDS (~ 3 years). We analyzed a large data set comprising ~ 350 pol and env sequences sampled in Cuba over the last 15 years and ~ 350 from Los Alamos database. This data set contained both CRF19 (~ 315), and A1, D and G sequences. We performed and combined analyses for the three A1, G and D regions, using fast maximum likelihood approaches, including: (1) phylogeny reconstruction, (2) spatio-temporal analysis of the virus spread, and ancestral character reconstruction for (3) transmission mode and (4) drug resistance mutations (DRMs). This allowed us to acquire new insights on the CRF19 origin and transmission patterns. We showed that CRF19 recombined between 1966 and 1977, most likely in Cuban community stationed in Congo region. We further investigated CRF19 spread on the Cuban province level, and discovered that the epidemic started in 1970s, most probably in Villa Clara, that it was at first carried by heterosexual transmissions, and then quickly spread in the 1980s within the "men having sex with men" (MSM) community, with multiple transmissions back to heterosexuals. The analysis of the transmission patterns of common DRMs showed mostly acquired drug resistance rather than transmitted one. Our results show a very early introduction of CRF19 in Cuba, which could explain its local epidemiological success. Ignited by a major founder event, the epidemic then followed a similar pattern as other subtypes and CRFs in Cuba. The reason for the short time to AIDS remains to be understood and requires specific surveillance, in Cuba and elsewhere. Author summaryCRF19 is a recombinant form of HIV-1, which causes rapid progression to AIDS (~ 3 years versus 5 - 10 years for other subtypes and CRFs). CRF19 is reported almost uniquely in Cuba, where it is highly prevalent (~ 25%) among newly detected HIV-1 patients. In this study, we found that CRF19 most likely recombined around the 1970s in the Cuban community that was stationed in Democratic Republic of the Congo and Angola at that time. It was introduced very early into the Cuban province of Villa Clara, from where it had several introductions to La Habana in the 1980s and then further spread to other Cuban provinces. The CRF19 epidemic most probably started with heterosexual transmissions, followed in the 1980s by multiple introductions into "men having sex with men" (MSM) community, followed by multiple transmissions back to heterosexuals (often females). The early introduction of CRF19 into Cuba most likely explains its success, not observed in other parts of the world. However, importantly, its rapid progression to AIDS makes it crucial to survey CRF19 sub-epidemics not only in Cuba, but also in other parts of the world having regular exchanges with Cuba.

bioinformatics↗

bioRxiv2021

Experimental and mathematical insights on the interactions between poliovirus and a defective interfering genome

1During replication, RNA viruses accumulate genome alterations, such as mutations and deletions. The interactions between individual variants can determine the fitness of the virus population and, thus, the outcome of infection. To investigate the effects of defective interfering genomes (DI) on wild-type (WT) poliovirus replication, we developed an ordinary differential equation model. We experimentally determined virus and DI replication during co-infection, and use these data to infer model parameters. Our model predicts, and our experimental measurements confirm, that DI replication and encapsidation are the most important determinants for the outcome of the competition. WT replication inversely correlates with DI replication. Our model predicts that genome replication and effective DI genome encapsidation are critical to effectively inhibit WT production, but an equilibrium can be established which enables WT to replicate, albeit to reduce levels.

microbiology↗

bioRxiv2020

COVID-Align: Accurate online alignment of hCoV-19 genomes using a profile HMM

MotivationThe first cases of the COVID-19 pandemic emerged in December 2019. Until the end of February 2020, the number of available genomes was below 1,000, and their multiple alignment was easily achieved using standard approaches. Subsequently, the availability of genomes has grown dramatically. Moreover, some genomes are of low quality with sequencing/assembly errors, making accurate re-alignment of all genomes nearly impossible on a daily basis. A more efficient, yet accurate approach was clearly required to pursue all subsequent bioinformatics analyses of this crucial data. ResultshCoV-19 genomes are highly conserved, with very few indels and no recombination. This makes the profile HMM approach particularly well suited to align new genomes, add them to an existing alignment and filter problematic ones. Using a core of [~]2,500 high quality genomes, we estimated a profile using HMMER, and implemented this profile in COVID-Align, a user-friendly interface to be used online or as standalone via Docker. The alignment of 1,000 genomes requires less than 20mn on our cluster. Moreover, COVID-Align provides summary statistics, which can be used to determine the sequencing quality and evolutionary novelty of input genomes (e.g. number of new mutations and indels). Availabilityhttps://covalign.pasteur.cloud, hub.docker.com/r/evolbioinfo/covid-align Contactsolivier.gascuel@pasteur.fr, frederic.lemoine@pasteur.fr Supplementary informationSupplementary information is available at Bioinformatics online.

bioinformatics↗

bioRxiv2019

Experimental and mathematical insights on the competition between poliovirus and a defective interfering genome

During replication, RNA virus populations accumulate genome alterations, such as mutations and deletions. The interactions between individual variants within the population can determine the fitness of the virus and, thus, the outcome of infection. We developed an ordinary differential equation model to infer the effect of the interaction between defective interfering (DI) replicons and wild-type (WT) poliovirus. We measure production of RNA and viral particles during a single infection cycle, and use these data to infer model parameters. We find that DI replicates faster than WT, but an equilibrium is established when both WT and DI compete for resources needed for RNA replication and genome encapsidation. In the presence of DI, the concentration of WT virions at cell lysis is suppressed by the factor of 5. Multiple generations within a single cell infection provide opportunities for significant inhibition of WT replication by competition with the faster replicating DI genomes.

microbiology↗

bioRxiv2017

Identifying Parameters Of Host Cell Vulnerability During Salmonella Infection By Quantitative Image Analysis And Modeling

Salmonella target and enter epithelial cells at permissive entry sites: some cells are more likely to be infected than others. However the parameters that lead to host cell heterogeneity are not known. Here, we quantitatively characterized host cell \"vulnerability\" towards Salmonella infection based on imaged parameters. We performed successive infections of the same host cell population followed by automated high-throuput microscopy and observed that infected cells have higher probability of being re-infected. Establishing a predictive model we identified two combined origins of host cell vulnerability: the pathogen-induced cellular vulnerability emerging from Salmonella uptake and persisting at later stage of the infection, and the host cell-inherent vulnerability. We linked the host cell inherent vulnerability with its morphological attributes such as the local cell crowding, and with host cell cholesterol content. This showed that the probability of Salmonella infection success can be forecast from morphological or molecular host cell parameters.

microbiology↗