{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,19]],"date-time":"2026-10-19T16:07:00Z","timestamp":1792426020158,"version":"4.5.0"},"reference-count":58,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,1,7]],"date-time":"2025-01-07T00:00:00Z","timestamp":1736208000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Med."],"abstract":"<jats:sec><jats:title>Background<\/jats:title><jats:p>Polygenic risk score (PRS) prediction is widely used to assess the risk of diagnosis and progression of many diseases. Routinely, the weights of individual SNPs are estimated by the linear regression model that assumes independent and linear contribution of each SNP to the phenotype. However, for complex multifactorial diseases such as Alzheimer\u2019s disease, diabetes, cardiovascular disease, cancer, and others, association between individual SNPs and disease could be non-linear due to epistatic interactions. The aim of the presented study is to explore the power of non-linear machine learning algorithms and deep learning models to predict the risk of multifactorial diseases with epistasis.<\/jats:p><\/jats:sec><jats:sec><jats:title>Methods<\/jats:title><jats:p>Simulated data with 2- and 3-loci interactions and tested three different models of epistasis: additive, multiplicative and threshold, were generated using the GAMETES. Penetrance tables were generated using PyTOXO package. For machine learning methods we used multilayer perceptron (MLP), convolutional neural network (CNN) and recurrent neural network (RNN), Lasso regression, random forest and gradient boosting models. Performance of machine learning models were assessed using accuracy, AUC-ROC, AUC-PR, recall, precision, and F1 score.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>First, we tested ensemble tree methods and deep learning neural networks against LASSO linear regression model on simulated data with different types and strength of epistasis. The results showed that with the increase of strength of epistasis effect, non-linear models significantly outperform linear. Then the higher performance of non-linear models over linear was confirmed on real genetic data for multifactorial phenotypes such as obesity, type 1 diabetes, and psoriasis. From non-linear models, gradient boosting appeared to be the best model in obesity and psoriasis while deep learning methods significantly outperform linear approaches in type 1 diabetes.<\/jats:p><\/jats:sec><jats:sec><jats:title>Conclusion<\/jats:title><jats:p>Overall, our study underscores the efficacy of non-linear models and deep learning approaches in more accurately accounting for the effects of epistasis in simulations with specific configurations and in the context of certain diseases.<\/jats:p><\/jats:sec>","DOI":"10.3389\/fmed.2024.1479717","type":"journal-article","created":{"date-parts":[[2025,1,7]],"date-time":"2025-01-07T05:34:45Z","timestamp":1736228085000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":12,"title":["Deep learning captures the effect of epistasis in multifactorial diseases"],"prefix":"10.3389","volume":"11","author":[{"given":"Vladislav","family":"Perelygin","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alexey","family":"Kamelin","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nikita","family":"Syzrantsev","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Layal","family":"Shaheen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Anna","family":"Kim","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nikolay","family":"Plotnikov","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Anna","family":"Ilinskaya","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Valery","family":"Ilinsky","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alexander","family":"Rakitko","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Maria","family":"Poptsova","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2025,1,7]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"267","DOI":"10.3389\/fgene.2019.00267","article-title":"Machine learning Snp based prediction for precision medicine","volume":"10","author":"Ho","year":"2019","journal-title":"Front Genet"},{"key":"ref2","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1016\/j.ajhg.2017.06.005","article-title":"10 years of Gwas discovery: biology, function, and translation","volume":"101","author":"Visscher","year":"2017","journal-title":"Am J Hum Genet"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"399","DOI":"10.1017\/S0080456800012163","article-title":"Xv.\u2014the correlation between relatives on the supposition of Mendelian inheritance","volume":"52","author":"Fisher","year":"2012","journal-title":"Earth Environ Sci Trans R Soc Edinb"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1101\/2021.03.17.21253794","article-title":"Atlas of epistasis","author":"Cl\u00e9ment","year":"2021","journal-title":"medRxiv"},{"key":"ref5","doi-asserted-by":"publisher","first-page":"359","DOI":"10.1038\/s10038-020-00832-7","article-title":"Deep neural network improves the estimation of polygenic risk scores for breast Cancer","volume":"66","author":"Badr\u00e9","year":"2021","journal-title":"J Hum Genet"},{"key":"ref6","doi-asserted-by":"publisher","first-page":"374","DOI":"10.1186\/s12885-022-09425-3","article-title":"Development and validation of polygenic risk scores for prediction of breast Cancer and breast Cancer subtypes in Chinese women","volume":"22","author":"Hou","year":"2022","journal-title":"BMC Cancer"},{"key":"ref7","doi-asserted-by":"publisher","first-page":"e84","DOI":"10.1016\/S2589-7500(21)00249-1","article-title":"Neural network-based integration of polygenic and clinical information: development and validation of a prediction model for 10-year risk of major adverse cardiac events in the Uk biobank cohort","volume":"4","author":"Steinfeldt","year":"2022","journal-title":"Lancet Digital Health"},{"key":"ref8","doi-asserted-by":"publisher","first-page":"e1010328","DOI":"10.1371\/journal.pcbi.1010328","article-title":"Explainable deep transfer learning model for disease risk prediction using high-dimensional genomic data","volume":"18","author":"Liu","year":"2022","journal-title":"PLoS Comput Biol"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"338","DOI":"10.1016\/j.ajhg.2018.07.015","article-title":"A one-penny imputed genome from next-generation reference panels","volume":"103","author":"Browning","year":"2018","journal-title":"Am J Hum Genet"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"241","DOI":"10.1038\/s41467-021-27930-0","article-title":"Deepnull models non-linear covariate effects to improve phenotypic prediction and association power","volume":"13","author":"McCaw","year":"2022","journal-title":"Nat Commun"},{"key":"ref11","doi-asserted-by":"publisher","first-page":"e0273293","DOI":"10.1371\/journal.pone.0273293","article-title":"Human genotype-to-phenotype predictions: boosting accuracy with nonlinear models","volume":"17","author":"Medvedev","year":"2022","journal-title":"PLoS One"},{"key":"ref12","year":"2023"},{"key":"ref13","doi-asserted-by":"publisher","first-page":"S3","DOI":"10.1002\/2327-6924.12510","article-title":"Obesity: risk factors, complications, and strategies for sustainable long-term weight management","volume":"29","author":"Fruh","year":"2017","journal-title":"J Am Assoc Nurse Pract"},{"key":"ref14","doi-asserted-by":"publisher","first-page":"220","DOI":"10.1001\/archgenpsychiatry.2010.2","article-title":"Overweight, obesity, and depression: a systematic review and Meta-analysis of longitudinal studies","volume":"67","author":"Luppino","year":"2010","journal-title":"Arch Gen Psychiatry"},{"key":"ref15","doi-asserted-by":"publisher","first-page":"706978","DOI":"10.3389\/fendo.2021.706978","article-title":"Obesity: epidemiology, pathophysiology, and therapeutics","volume":"12","author":"Lin","year":"2021","journal-title":"Front Endocrinol"},{"key":"ref16","doi-asserted-by":"publisher","first-page":"29","DOI":"10.3389\/fendo.2012.00029","article-title":"Variability in the heritability of body mass index: a systematic review and Meta-regression","volume":"3","author":"Elks","year":"2012","journal-title":"Front Endocrinol"},{"key":"ref17","doi-asserted-by":"publisher","first-page":"669155","DOI":"10.3389\/fnut.2021.669155","article-title":"Predicting obesity in adults using machine learning techniques: an analysis of Indonesian basic Health Research 2018","volume":"8","author":"Thamrin","year":"2021","journal-title":"Front Nutr"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"e15992","DOI":"10.2196\/15992","article-title":"Deep learning neural networks to predict serious complications after bariatric surgery: analysis of Scandinavian obesity surgery registry data","volume":"8","author":"Cao","year":"2020","journal-title":"JMIR Med Inform"},{"key":"ref19","doi-asserted-by":"publisher","first-page":"506","DOI":"10.4338\/aci-2015-03-ra-0036","article-title":"Machine learning techniques for prediction of early childhood obesity","volume":"6","author":"Dugan","year":"2015","journal-title":"Appl Clin Inform"},{"key":"ref20","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3390\/s21072361","article-title":"The prediction of body mass index from negative affectivity through machine learning: a confirmatory study","volume":"21","author":"Delnevo","year":"2021","journal-title":"Sensors"},{"key":"ref21","doi-asserted-by":"publisher","first-page":"783845","DOI":"10.3389\/fgene.2021.783845","article-title":"Using machine learning to predict obesity based on genome-wide and epigenome-wide Gene-gene and Gene-diet interactions","volume":"12","author":"Lee","year":"2021","journal-title":"Front Genet"},{"key":"ref22","doi-asserted-by":"publisher","first-page":"1347","DOI":"10.1089\/cmb.2018.0002","article-title":"Machine learning-based method for obesity risk evaluation using single-nucleotide polymorphisms derived from next-generation sequencing","volume":"25","author":"Wang","year":"2018","journal-title":"J computational biol: J computational molecular cell biology"},{"key":"ref23","doi-asserted-by":"publisher","first-page":"88","DOI":"10.1016\/j.gene.2019.03.011","article-title":"Decision tree learning to predict overweight\/obesity based on body mass index and gene Polymporphisms","volume":"699","author":"Rodr\u00edguez-Pardo","year":"2019","journal-title":"Gene"},{"key":"ref24","doi-asserted-by":"publisher","first-page":"e0197843","DOI":"10.1371\/journal.pone.0197843","article-title":"A computational framework for predicting obesity risk based on optimizing and integrating genetic risk score and gene expression profiles","volume":"13","author":"Joseph","year":"2018","journal-title":"PLoS One"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3390\/biomedicines10102436","article-title":"Current advances in the Management of Diabetes Mellitus","volume":"10","author":"Aloke","year":"2022","journal-title":"Biomedicine"},{"key":"ref26","doi-asserted-by":"publisher","first-page":"1167","DOI":"10.1007\/s00125-019-4863-8","article-title":"Type 1 diabetes defined by severe insulin deficiency occurs after 30 years of age and is commonly treated as type 2 diabetes","volume":"62","author":"Thomas","year":"2019","journal-title":"Diabetologia"},{"key":"ref27","doi-asserted-by":"publisher","first-page":"104383","DOI":"10.1016\/j.ebiom.2022.104383","article-title":"Prediction of type 2 diabetes using genome-wide polygenic risk score and metabolic profiles: a machine learning analysis of population-based 10-year prospective cohort study","volume":"86","author":"Hahn","year":"2022","journal-title":"EBioMedicine"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3390\/diagnostics12123067","article-title":"Using recurrent neural networks for predicting Type-2 diabetes from genomic and tabular data","volume":"12","author":"Srinivasu","year":"2022","journal-title":"Diagnostics"},{"key":"ref29","doi-asserted-by":"publisher","first-page":"148","DOI":"10.1186\/s13098-021-00767-9","article-title":"Machine learning and deep learning predictive models for type 2 diabetes: a systematic review","volume":"13","author":"Fregoso-Aparicio","year":"2021","journal-title":"Diabetol Metab Syndr"},{"key":"ref30","doi-asserted-by":"publisher","first-page":"e0282426","DOI":"10.1371\/journal.pone.0282426","article-title":"Predicting the development of T1d and identifying its key performance indicators in children; a case-control study in Saudi Arabia","volume":"18","author":"Alazwari","year":"2023","journal-title":"PLoS One"},{"key":"ref31","doi-asserted-by":"publisher","first-page":"110029","DOI":"10.1016\/j.diabres.2022.110029","article-title":"Predicting misdiagnosed adult-onset type 1 diabetes using machine learning","volume":"191","author":"Cheheltani","year":"2022","journal-title":"Diabetes Res Clin Pract"},{"key":"ref32","doi-asserted-by":"publisher","first-page":"115648","DOI":"10.1016\/j.eswa.2021.115648","article-title":"Machine learning analysis of the human infant gut microbiome identifies influential species in type 1 diabetes","volume":"185","author":"Fern\u00e1ndez-Edreira","year":"2021","journal-title":"Expert Syst Appl"},{"key":"ref33","doi-asserted-by":"publisher","first-page":"185","DOI":"10.1007\/s00430-022-00741-w","article-title":"Child type 1 diabetes associated with mother vaginal Bacteriome and Mycobiome","volume":"211","author":"Ruotsalainen","year":"2022","journal-title":"Med Microbiol Immunol"},{"key":"ref34","doi-asserted-by":"publisher","first-page":"412","DOI":"10.1111\/bjd.12375","article-title":"Heritability of psoriasis in a large twin sample","volume":"169","author":"L\u00f8nnberg","year":"2013","journal-title":"Br J Dermatol"},{"key":"ref35","doi-asserted-by":"publisher","first-page":"2","DOI":"10.1016\/j.jdermsci.2020.05.008","article-title":"The current landscape of psoriasis genetics in 2020","volume":"99","author":"Ogawa","year":"2020","journal-title":"J Dermatol Sci"},{"key":"ref36","doi-asserted-by":"publisher","first-page":"378","DOI":"10.1111\/jocd.15122","article-title":"Therapeutic application of machine learning in psoriasis: a prisma systematic review","volume":"22","author":"Lunge","year":"2023","journal-title":"J Cosmet Dermatol"},{"key":"ref37","doi-asserted-by":"publisher","first-page":"381","DOI":"10.1038\/sj.gene.6364106","article-title":"Genotype effects and epistasis in type 1 diabetes and Hla-Dq trans dimer associations with disease","volume":"5","author":"Koeleman","year":"2004","journal-title":"Genes Immun"},{"key":"ref38","doi-asserted-by":"publisher","first-page":"117","DOI":"10.1186\/s12859-022-04645-7","article-title":"Pytoxo: a Python tool for calculating penetrance tables of high-order epistasis models","volume":"23","author":"Gonz\u00e1lez-Seoane","year":"2022","journal-title":"BMC Bioinfo"},{"key":"ref39","doi-asserted-by":"publisher","first-page":"16","DOI":"10.1186\/1756-0381-5-16","article-title":"Gametes: a fast, direct algorithm for generating pure, strict, epistatic models with random architectures","volume":"5","author":"Urbanowicz","year":"2012","journal-title":"BioData mining"},{"key":"ref40","year":""},{"key":"ref41","doi-asserted-by":"publisher","first-page":"192","DOI":"10.1038\/tpj.2013.18","article-title":"HIBAG--HLA genotype imputation with attribute bagging","volume":"14","author":"Zheng","year":"2014","journal-title":"Pharm J"},{"key":"ref42","doi-asserted-by":"publisher","first-page":"26","DOI":"10.1038\/s41588-017-0011-x","article-title":"Protein-altering variants associated with body mass index implicate pathways that control energy intake and expenditure in obesity","volume":"50","author":"Turcot","year":"2018","journal-title":"Nat Genet"},{"key":"ref43","doi-asserted-by":"publisher","first-page":"559","DOI":"10.1086\/519795","article-title":"Plink: a tool set for whole-genome association and population-based linkage analyses","volume":"81","author":"Purcell","year":"2007","journal-title":"Am J Hum Genet"},{"key":"ref44","doi-asserted-by":"publisher","first-page":"e0179348","DOI":"10.1371\/journal.pone.0179348","article-title":"The association between 38 previously reported polymorphisms and psoriasis in a polish population: high predicative accuracy of a genetic risk score combining 16 loci","volume":"12","author":"Kisiel","year":"2017","journal-title":"PLoS One"},{"key":"ref45","doi-asserted-by":"publisher","first-page":"200","DOI":"10.2337\/dc18-1785","article-title":"Development and standardization of an improved type 1 diabetes genetic risk score for use in newborn screening and incident diagnosis","volume":"42","author":"Sharp","year":"2019","journal-title":"Diabetes Care"},{"key":"ref46","doi-asserted-by":"publisher","first-page":"1112","DOI":"10.1093\/ejcts\/ezy167","article-title":"Statistical primer: propensity score matching and its alternatives","volume":"53","author":"Benedetto","year":"2018","journal-title":"European J cardio-thoracic Surg: Official J European Association for Cardio-thoracic Surg"},{"key":"ref47","doi-asserted-by":"publisher","first-page":"178","DOI":"10.1016\/j.amjmed.2019.08.055","article-title":"Propensity score matching: the 'Devil is in the Details' where more may be hidden than you know","volume":"133","author":"Reiffel","year":"2020","journal-title":"Am J Med"},{"key":"ref48","doi-asserted-by":"publisher","first-page":"1396","DOI":"10.1007\/s11606-020-05713-5","article-title":"Benchmarking observational analyses against randomized trials: a review of studies assessing propensity score methods","volume":"35","author":"Forbes","year":"2020","journal-title":"J Gen Intern Med"},{"key":"ref49","doi-asserted-by":"publisher","first-page":"413","DOI":"10.1038\/ng1537","article-title":"Genome-wide strategies for detecting multiple loci that influence complex diseases","volume":"37","author":"Marchini","year":"2005","journal-title":"Nat Genet"},{"key":"ref50","doi-asserted-by":"publisher","first-page":"809","DOI":"10.1534\/genetics.118.301298","article-title":"Can deep learning improve genomic prediction of complex human traits?","volume":"210","author":"Bellot","year":"2018","journal-title":"Genetics"},{"key":"ref51","doi-asserted-by":"publisher","first-page":"125","DOI":"10.1002\/gepi.22279","article-title":"Polygenic risk scores outperform machine learning methods in predicting coronary artery disease status","volume":"44","author":"Gola","year":"2020","journal-title":"Genet Epidemiol"},{"key":"ref52","doi-asserted-by":"publisher","first-page":"7306","DOI":"10.1038\/s41598-022-11270-0","article-title":"Concurrent outcomes from multiple approaches of epistasis analysis for human body mass index associated loci provide insights into obesity biology","volume":"12","author":"D'Silva","year":"2022","journal-title":"Sci Rep"},{"key":"ref53","doi-asserted-by":"publisher","first-page":"3641","DOI":"10.1093\/hmg\/ddy271","article-title":"Meta-analysis of genome-wide association studies for height and body mass index in \u223c700000 individuals of European ancestry","volume":"27","author":"Yengo","year":"2018","journal-title":"Hum Mol Genet"},{"key":"ref54","doi-asserted-by":"publisher","first-page":"e162","DOI":"10.1038\/cti.2017.51","article-title":"Type 1 diabetes genome-wide association studies: not to be lost in translation","volume":"6","author":"Pociot","year":"2017","journal-title":"Clinical & Translational Immunol"},{"key":"ref55","doi-asserted-by":"publisher","first-page":"2320","DOI":"10.1038\/jid.2012.167","article-title":"Genetics of psoriasis: evidence for epistatic interaction between skin barrier abnormalities and immune deviation","volume":"132","author":"Bergboer","year":"2012","journal-title":"J Invest Dermatol"},{"key":"ref56","doi-asserted-by":"publisher","first-page":"15186","DOI":"10.1038\/s41598-018-33493-w","article-title":"Discovering genetic factors for psoriasis through exhaustively searching for significant second order Snp-Snp interactions","volume":"8","author":"Lee","year":"2018","journal-title":"Sci Rep"},{"key":"ref57","doi-asserted-by":"publisher","first-page":"e0163185","DOI":"10.1371\/journal.pone.0163185","article-title":"Evaluation of psoriasis genetic risk based on five susceptibility markers in a population from northern Poland","volume":"11","author":"Stawczyk-Macieja","year":"2016","journal-title":"PLoS One"},{"key":"ref58","doi-asserted-by":"publisher","first-page":"e19454","DOI":"10.1371\/journal.pone.0019454","article-title":"A genetic risk score combining ten psoriasis risk loci improves disease prediction","volume":"6","author":"Chen","year":"2011","journal-title":"PLoS One"}],"container-title":["Frontiers in Medicine"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2024.1479717\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,1,7]],"date-time":"2025-01-07T05:34:49Z","timestamp":1736228089000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2024.1479717\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,1,7]]},"references-count":58,"alternative-id":["10.3389\/fmed.2024.1479717"],"URL":"https:\/\/doi.org\/10.3389\/fmed.2024.1479717","relation":{},"ISSN":["2296-858X"],"issn-type":[{"value":"2296-858X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,1,7]]},"article-number":"1479717"}}