Genotype-to-phenotype prediction is a central problem of human genetics. In recent years, it has become possible to construct complex predictive models for phenotypes, thanks to the availability of large genome data sets as well as efficient and scalable machine learning tools. In this paper, we make a threefold contribution to this problem. First, we ask if state-of-the-art nonlinear predictive models, such as boosted decision trees, can be more efficient for phenotype prediction than conventional linear models. We find that this is indeed the case if model features include a sufficiently rich set of covariates, but probably not otherwise. Second, we ask if the conventional selection of single nucleotide polymorphisms (SNPs) by genome wide association studies (GWAS) can be replaced by a more efficient procedure, taking into account information in previously selected SNPs. We propose such a procedure, based on a sequential feature importance estimation with decision trees, and show that this approach indeed produced informative SNP sets that are much more compact than when selected with GWAS. Finally, we show that the highest prediction accuracy can ultimately be achieved by ensembling individual linear and nonlinear models. To the best of our knowledge, for some of the phenotypes that we consider (asthma, hypothyroidism), our results are a new state-of-the-art.

Combining training data from multiple sources increases sample size and reduces confounding, leading to more accurate and less biased machine learning models. In healthcare, however, direct pooling of data is often not allowed by data custodians who are accountable for minimizing the exposure of sensitive information. Federated learning offers a promising solution to this problem by training a model in a decentralized manner thus reducing the risks of data leakage. Although there is increasing utilization of federated learning on clinical data, its efficacy on genomic data has not been extensively studied. This study aims to contribute to the adoption of federated learning for genomic data by investigating its applicability in two scenarios: phenotype prediction on the UK Biobank data and ancestry prediction on the 1000 Genomes Project data. By splitting data into independent datasets and implementing different federated learning strategies, we show that federated models achieve performance close to centralized models, even in the presence of significant inter-node heterogeneity. This paper describes the experiments and provides recommendations on strategies that should be used to reduce computational complexity or communication costs.

AbstractHigh-throughput sequencing of fetal DNA is a promising and increasingly common method for the discovery of all (or all coding) genetic variants in the fetus, either as part of prenatal screening or diagnosis, or for genetic diagnosis of spontaneous abortions. In many cases, the fetal DNA (from chorionic villi, amniotic fluid, or abortive tissue) can be contaminated with maternal cells, resulting in the mixture of fetal and maternal DNA. This maternal cell contamination (MCC) undermines the assumption, made by traditional variant callers, that each allele in a heterozygous site is covered, on average, by 50% of the reads, and therefore can lead to erroneous genotype calls. We present a panel of methods for reducing the genotyping error in the presence of MCC. All methods start with the output of GATK HaplotypeCaller on the sequencing data for the (contaminated) fetal sample and both of its parents, and additionally rely on information about the MCC fraction (which itself is readily estimated from the high-throughput sequencing data). The first of these methods uses a Bayesian probabilistic model to correct the fetal genotype calls produced by MCC-unaware HaplotypeCaller. The other two methods “learn” the genotype-correction model from examples. We use simulated contaminated fetal data to train and test the models. Using the test sets, we show that all three methods lead to substantially improved accuracy when compared with the original MCC-unaware HaplotypeCaller calls. We then apply the best-performing method to three chorionic villus samples from spontaneously terminated pregnancies.Code and training data availabilityhttps://github.com/bazykinlab/ML-maternal-cell-contamination

High-throughput sequencing of fetal DNA is a promising and increasingly common method for the discovery of all (or all coding) genetic variants in the fetus, either as part of prenatal screening or diagnosis, or for genetic diagnosis of spontaneous abortions. In many cases, the fetal DNA (from chorionic villi, amniotic fluid, or abortive tissue) can be contaminated with maternal cells, resulting in the mixture of fetal and maternal DNA. This maternal cell contamination (MCC) undermines the assumption, made by traditional variant callers, that each allele in a heterozygous site is covered, on average, by 50% of the reads, and therefore can lead to erroneous genotype calls. We present a panel of methods for reducing the genotyping error in the presence of MCC. All methods start with the output of GATK HaplotypeCaller on the sequencing data for the (contaminated) fetal sample and both of its parents, and additionally rely on information about the MCC fraction (which itself is readily estimated from the high-throughput sequencing data). The first of these methods uses a Bayesian probabilistic model to correct the fetal genotype calls produced by MCC-unaware HaplotypeCaller. The other two methods "learn" the genotype-correction model from examples. We use simulated contaminated fetal data to train and test the models. Using the test sets, we show that all three methods lead to substantially improved accuracy when compared with the original MCCunaware HaplotypeCaller calls. We then apply the best-performing method to three chorionic villus samples from spontaneously terminated pregnancies.

Prediction of protein stability change due to single mutation is important for biotechnology, medicine, and our understanding of physics underlying protein folding. Despite the recent tremendous success in 3D protein structure prediction, the apparently simpler problem of predicting the effect of mutations on protein stability has been hampered by the low amount of experimental data. With the recent high-throughput measurements of mutational effects in 'mega' experiment for ~850,000 mutations [Tsuboyama et al., bioRxiv, 2022] it becomes possible to apply the state-of-the-art deep learning methods. Here we explore the ability of ESM2 deep neural network architecture with added Light Attention mechanism to predict the change of protein stability due to single mutations. The resulting method ABYSSAL predicts well the data from the 'mega' experiment (Pearson correlation 0.85) while the prediction of DDG values from previous experiments is more modest (Pearson correlation 0.50). ABYSSAL also shows a perfect satisfaction of the antisymmetry property. The ABYSSAL training demonstrated that the dataset should contain around ~100,000 data points for taking advantage of the state-of-the-art deep learning methods. Overall, our study shows great perspectives for developing the deep learning DDG predictors.