{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,28]],"date-time":"2026-03-28T04:23:59Z","timestamp":1774671839889,"version":"3.50.1"},"update-to":[{"DOI":"10.1371\/journal.pcbi.1010050","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2022,4,21]],"date-time":"2022-04-21T00:00:00Z","timestamp":1650499200000}}],"reference-count":51,"publisher":"Public Library of Science (PLoS)","issue":"4","license":[{"start":{"date-parts":[[2022,4,11]],"date-time":"2022-04-11T00:00:00Z","timestamp":1649635200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>Scientific research is shedding light on the interaction of the gut microbiome with the human host and on its role in human health. Existing machine learning methods have shown great potential in discriminating healthy from diseased microbiome states. Most of them leverage shotgun metagenomic sequencing to extract gut microbial species-relative abundances or strain-level markers. Each of these gut microbial profiling modalities showed diagnostic potential when tested separately; however, no existing approach combines them in a single predictive framework. Here, we propose the Multimodal Variational Information Bottleneck (MVIB), a novel deep learning model capable of learning a joint representation of multiple heterogeneous data modalities. MVIB achieves competitive classification performance while being faster than existing methods. Additionally, MVIB offers interpretable results. Our model adopts an information theoretic interpretation of deep neural networks and computes a joint stochastic encoding of different input data modalities. We use MVIB to predict whether human hosts are affected by a certain disease by jointly analysing gut microbial species-relative abundances and strain-level markers. MVIB is evaluated on human gut metagenomic samples from 11 publicly available disease cohorts covering 6 different diseases. We achieve high performance (0.80 &lt; ROC AUC &lt; 0.95) on 5 cohorts and at least medium performance on the remaining ones. We adopt a saliency technique to interpret the output of MVIB and identify the most relevant microbial species and strain-level markers to the model\u2019s predictions. We also perform cross-study generalisation experiments, where we train and test MVIB on different cohorts of the same disease, and overall we achieve comparable results to the baseline approach, i.e. the Random Forest. Further, we evaluate our model by adding metabolomic data derived from mass spectrometry as a third input modality. Our method is scalable with respect to input data modalities and has an average training time of &lt; 1.4 seconds. The source code and the datasets used in this work are publicly available.<\/jats:p>","DOI":"10.1371\/journal.pcbi.1010050","type":"journal-article","created":{"date-parts":[[2022,4,11]],"date-time":"2022-04-11T13:35:49Z","timestamp":1649684149000},"page":"e1010050","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":37,"title":["Microbiome-based disease prediction with multimodal variational information bottlenecks"],"prefix":"10.1371","volume":"18","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-8888-133X","authenticated-orcid":true,"given":"Filippo","family":"Grazioli","sequence":"first","affiliation":[]},{"given":"Raman","family":"Siarheyeu","sequence":"additional","affiliation":[]},{"given":"Israa","family":"Alqassem","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1386-5372","authenticated-orcid":true,"given":"Andreas","family":"Henschel","sequence":"additional","affiliation":[]},{"given":"Giampaolo","family":"Pileggi","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9666-8456","authenticated-orcid":true,"given":"Andrea","family":"Meiser","sequence":"additional","affiliation":[]}],"member":"340","published-online":{"date-parts":[[2022,4,11]]},"reference":[{"issue":"2","key":"pcbi.1010050.ref001","doi-asserted-by":"crossref","first-page":"210","DOI":"10.1016\/j.chom.2019.01.018","article-title":"The battle within: interactions of bacteriophages and bacteria in the gastrointestinal tract","volume":"25","author":"L De Sordi","year":"2019","journal-title":"Cell host and microbe"},{"issue":"4","key":"pcbi.1010050.ref002","doi-asserted-by":"crossref","first-page":"392","DOI":"10.1038\/nm.4517","article-title":"Current understanding of the human microbiome","volume":"24","author":"JA Gilbert","year":"2018","journal-title":"Nature medicine"},{"issue":"4","key":"pcbi.1010050.ref003","doi-asserted-by":"crossref","first-page":"260","DOI":"10.1038\/nrg3182","article-title":"The human microbiome: at the interface of health and disease","volume":"13","author":"I Cho","year":"2012","journal-title":"Nature Reviews Genetics"},{"issue":"7402","key":"pcbi.1010050.ref004","doi-asserted-by":"crossref","first-page":"207","DOI":"10.1038\/nature11234","article-title":"Structure, function and diversity of the healthy human microbiome","volume":"486","author":"C Huttenhower","year":"2012","journal-title":"Nature"},{"issue":"2","key":"pcbi.1010050.ref005","doi-asserted-by":"crossref","first-page":"e77","DOI":"10.1016\/S1470-2045(18)30952-5","article-title":"Modulating the microbiome to improve therapeutic response in cancer","volume":"20","author":"JL McQuade","year":"2019","journal-title":"The Lancet Oncology"},{"key":"pcbi.1010050.ref006","doi-asserted-by":"crossref","first-page":"145","DOI":"10.1146\/annurev-med-010312-133513","article-title":"The human microbiome: from symbiosis to pathogenesis","volume":"64","author":"EA Eloe-Fadrosh","year":"2013","journal-title":"Annual review of medicine"},{"issue":"7164","key":"pcbi.1010050.ref007","doi-asserted-by":"crossref","first-page":"804","DOI":"10.1038\/nature06244","article-title":"The human microbiome project","volume":"449","author":"PJ Turnbaugh","year":"2007","journal-title":"Nature"},{"issue":"7285","key":"pcbi.1010050.ref008","doi-asserted-by":"crossref","first-page":"59","DOI":"10.1038\/nature08821","article-title":"A human gut microbial gene catalogue established by metagenomic sequencing","volume":"464","author":"J Qin","year":"2010","journal-title":"Nature"},{"key":"pcbi.1010050.ref009","doi-asserted-by":"crossref","first-page":"277","DOI":"10.3389\/fmicb.2021.635781","article-title":"Statistical and machine learning techniques in human microbiome studies: contemporary challenges and solutions","volume":"12","author":"I Moreno-Indias","year":"2021","journal-title":"Frontiers in Microbiology"},{"issue":"9","key":"pcbi.1010050.ref010","doi-asserted-by":"crossref","first-page":"833","DOI":"10.1038\/nbt.3935","article-title":"Shotgun metagenomics, from sampling to analysis","volume":"35","author":"C Quince","year":"2017","journal-title":"Nature biotechnology"},{"issue":"7","key":"pcbi.1010050.ref011","doi-asserted-by":"crossref","first-page":"410","DOI":"10.1038\/s41579-018-0029-9","article-title":"Best practices for analysing microbiomes","volume":"16","author":"R Knight","year":"2018","journal-title":"Nature Reviews Microbiology"},{"issue":"7","key":"pcbi.1010050.ref012","doi-asserted-by":"crossref","first-page":"e1004977","DOI":"10.1371\/journal.pcbi.1004977","article-title":"Machine learning meta-analysis of large metagenomic datasets: tools and biological insights","volume":"12","author":"E Pasolli","year":"2016","journal-title":"PLoS computational biology"},{"issue":"4","key":"pcbi.1010050.ref013","doi-asserted-by":"crossref","first-page":"626","DOI":"10.1101\/gr.216242.116","article-title":"Microbial strain-level population structure and genetic diversity from metagenomes","volume":"27","author":"DT Truong","year":"2017","journal-title":"Genome research"},{"key":"pcbi.1010050.ref014","unstructured":"Nguyen TH, Chevaleyre Y, Prifti E, Sokolovska N, Zucker JD. Deep learning for metagenomic data: using 2d embeddings and convolutional neural networks. arXiv:171200244 [Preprint]. 2017 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1712.00244"},{"key":"pcbi.1010050.ref015","unstructured":"Nguyen TH, Prifti E, Chevaleyre Y, Sokolovska N, Zucker JD. Disease classification in metagenomics with 2d embeddings and deep learning. arXiv:180609046 [Preprint]. 2018 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1806.09046"},{"issue":"13","key":"pcbi.1010050.ref016","doi-asserted-by":"crossref","first-page":"i32","DOI":"10.1093\/bioinformatics\/bty296","article-title":"MicroPheno: predicting environments and host phenotypes from 16S rRNA gene sequencing using a k-mer based representation of shallow sub-samples","volume":"34","author":"E Asgari","year":"2018","journal-title":"Bioinformatics"},{"issue":"1","key":"pcbi.1010050.ref017","first-page":"1","article-title":"DeepMicro: deep representation learning for disease prediction based on microbiome data","volume":"10","author":"M Oh","year":"2020","journal-title":"Scientific reports"},{"issue":"5","key":"pcbi.1010050.ref018","doi-asserted-by":"crossref","first-page":"1555","DOI":"10.1161\/HYPERTENSIONAHA.120.15885","article-title":"Machine Learning Strategy for Gut Microbiome-Based Diagnostic Screening of Cardiovascular Disease","volume":"76","author":"S Aryal","year":"2020","journal-title":"Hypertension"},{"issue":"10","key":"pcbi.1010050.ref019","doi-asserted-by":"crossref","first-page":"2993","DOI":"10.1109\/JBHI.2020.2993761","article-title":"PopPhy-CNN: a phylogenetic tree embedded architecture for convolutional neural networks to predict host phenotype from metagenomic data","volume":"24","author":"D Reiman","year":"2020","journal-title":"IEEE journal of biomedical and health informatics"},{"issue":"1","key":"pcbi.1010050.ref020","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1186\/s13059-021-02306-1","article-title":"Microbiome meta-analysis and cross-disease comparison enabled by the SIAMCAT machine learning toolbox","volume":"22","author":"J Wirbel","year":"2021","journal-title":"Genome Biology"},{"key":"pcbi.1010050.ref021","doi-asserted-by":"crossref","first-page":"627","DOI":"10.3389\/fgene.2019.00627","article-title":"Multitable Methods for Microbiome Data Integration","volume":"10","author":"K Sankaran","year":"2019","journal-title":"Frontiers in Genetics"},{"issue":"11","key":"pcbi.1010050.ref022","doi-asserted-by":"crossref","first-page":"1112","DOI":"10.1158\/1940-6207.CAPR-14-0129","article-title":"The human gut microbiome as a screening tool for colorectal cancer","volume":"7","author":"JP Zackular","year":"2014","journal-title":"Cancer prevention research"},{"key":"pcbi.1010050.ref023","unstructured":"Alemi AA, Fischer I, Dillon JV, Murphy K. Deep variational information bottleneck. arXiv:161200410 [Preprint]. 2016 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1612.00410"},{"key":"pcbi.1010050.ref024","unstructured":"Tishby N, Pereira FC, Bialek W. The information bottleneck method. arXiv:0004057 [Preprint]. 2000 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/physics\/0004057"},{"key":"pcbi.1010050.ref025","unstructured":"Simonyan K, Vedaldi A, Zisserman A. Deep inside convolutional networks: Visualising image classification models and saliency maps. arXiv:13126034 [Preprint]. 2013 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1312.6034"},{"issue":"10","key":"pcbi.1010050.ref026","doi-asserted-by":"crossref","first-page":"1345","DOI":"10.1109\/TKDE.2009.191","article-title":"A survey on transfer learning","volume":"22","author":"SJ Pan","year":"2009","journal-title":"IEEE Transactions on knowledge and data engineering"},{"issue":"1","key":"pcbi.1010050.ref027","doi-asserted-by":"crossref","first-page":"43","DOI":"10.1109\/JPROC.2020.3004555","article-title":"A comprehensive survey on transfer learning","volume":"109","author":"F Zhuang","year":"2020","journal-title":"Proceedings of the IEEE"},{"key":"pcbi.1010050.ref028","unstructured":"Kingma DP, Welling M. Auto-encoding variational bayes. arXiv:13126114 [Preprint]. 2013 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1312.6114"},{"issue":"7452","key":"pcbi.1010050.ref029","doi-asserted-by":"crossref","first-page":"99","DOI":"10.1038\/nature12198","article-title":"Gut metagenome in European women with normal, impaired and diabetic glucose control","volume":"498","author":"FH Karlsson","year":"2013","journal-title":"Nature"},{"issue":"7418","key":"pcbi.1010050.ref030","doi-asserted-by":"crossref","first-page":"55","DOI":"10.1038\/nature11450","article-title":"A metagenome-wide association study of gut microbiota in type 2 diabetes","volume":"490","author":"J Qin","year":"2012","journal-title":"Nature"},{"issue":"7464","key":"pcbi.1010050.ref031","doi-asserted-by":"crossref","first-page":"541","DOI":"10.1038\/nature12506","article-title":"Richness of human gut microbiome correlates with metabolic markers","volume":"500","author":"E Le Chatelier","year":"2013","journal-title":"Nature"},{"issue":"6285","key":"pcbi.1010050.ref032","doi-asserted-by":"crossref","first-page":"586","DOI":"10.1126\/science.aad8852","article-title":"Durable coexistence of donor and recipient strains after fecal microbiota transplantation","volume":"352","author":"S Li","year":"2016","journal-title":"Science"},{"issue":"73","key":"pcbi.1010050.ref033","article-title":"Temporal and technical variability of human gut metagenomes","volume":"16","author":"A Voigt","year":"2015","journal-title":"Genome Biology"},{"issue":"7516","key":"pcbi.1010050.ref034","doi-asserted-by":"crossref","first-page":"59","DOI":"10.1038\/nature13568","article-title":"Alterations of the human gut microbiome in liver cirrhosis","volume":"513","author":"N Qin","year":"2014","journal-title":"Nature"},{"issue":"11","key":"pcbi.1010050.ref035","doi-asserted-by":"crossref","first-page":"766","DOI":"10.15252\/msb.20145645","article-title":"Potential of fecal microbiota for early-stage detection of colorectal cancer","volume":"10","author":"G Zeller","year":"2014","journal-title":"Molecular systems biology"},{"issue":"14","key":"pcbi.1010050.ref036","article-title":"Gut microbiota dysbiosis contributes to the development of hypertension","volume":"5","author":"J Li","year":"2017","journal-title":"Microbiome"},{"issue":"6","key":"pcbi.1010050.ref037","doi-asserted-by":"crossref","first-page":"968","DOI":"10.1038\/s41591-019-0458-7","article-title":"Metagenomic and metabolomic analyses reveal distinct stage-specific phenotypes of the gut microbiota in colorectal cancer","volume":"25","author":"S Yachida","year":"2019","journal-title":"Nature Medicine"},{"key":"pcbi.1010050.ref038","unstructured":"NCBI. SRA-Tools. Last visited 2021 May 21. Available from: http:\/\/ncbi.github.io\/sra-tools\/"},{"key":"pcbi.1010050.ref039","unstructured":"Biobakery. KneadData. Last visited 2021 May 21. Available from: https:\/\/github.com\/biobakery\/kneaddata"},{"issue":"10","key":"pcbi.1010050.ref040","doi-asserted-by":"crossref","first-page":"902","DOI":"10.1038\/nmeth.3589","article-title":"MetaPhlAn2 for enhanced metagenomic taxonomic profiling","volume":"12","author":"DT Truong","year":"2015","journal-title":"Nature methods"},{"key":"pcbi.1010050.ref041","unstructured":"Wu M, Goodman N. Multimodal generative models for scalable weakly-supervised learning. arXiv:180205335 [Preprint]. 2018 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1802.05335"},{"key":"pcbi.1010050.ref042","unstructured":"Cao Y, Fleet DJ. Generalized product of experts for automatic and principled fusion of Gaussian process predictions. arXiv:14107827 [Preprint]. 2014 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1410.7827"},{"key":"pcbi.1010050.ref043","unstructured":"Hendrycks D, Gimpel K. Gaussian error linear units (gelus). arXiv:160608415 [Preprint]. 2016 [cited 2021 May 21]. Available from: https:\/\/arxiv.org\/abs\/1606.08415"},{"key":"pcbi.1010050.ref044","unstructured":"Paszke A, Gross S, Massa F, Lerer A, Bradbury J, et al. PyTorch: An Imperative Style, High-Performance Deep Learning Library Advances in Neural Information Processing Systems. 2019. 8024\u20138035"},{"key":"pcbi.1010050.ref045","article-title":"Learning local feature descriptors with triplets and shallow convolutional neural networks","author":"V Balntas","year":"2016","journal-title":"BMVC"},{"key":"pcbi.1010050.ref046","first-page":"2825","article-title":"Scikit-learn: Machine Learning in Python","volume":"12","author":"F Pedregosa","year":"2011","journal-title":"Journal of Machine Learning Research"},{"issue":"1913","key":"pcbi.1010050.ref047","first-page":"2019","article-title":"It\u2019s what\u2019s on the inside that counts: stress physiology and the bacterial microbiome of a wild urban mammal","volume":"286","author":"MR Stothart","year":"2019","journal-title":"Proceedings of the Royal Society B"},{"issue":"10","key":"pcbi.1010050.ref048","doi-asserted-by":"crossref","first-page":"e0222394","DOI":"10.1371\/journal.pone.0222394","article-title":"Gut microbiome diversity is associated with sleep physiology in humans","volume":"14","author":"RP Smith","year":"2019","journal-title":"PLoS One"},{"issue":"Supplement 1","key":"pcbi.1010050.ref049","doi-asserted-by":"crossref","first-page":"4680","DOI":"10.1073\/pnas.1002611107","article-title":"Vaginal microbiome of reproductive-age women","volume":"108","author":"J Ravel","year":"2011","journal-title":"Proceedings of the National Academy of Sciences"},{"issue":"1","key":"pcbi.1010050.ref050","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1038\/ncomms4654","article-title":"Gut microbiome of the Hadza hunter-gatherers","volume":"5","author":"SL Schnorr","year":"2014","journal-title":"Nature communications"},{"issue":"8","key":"pcbi.1010050.ref051","doi-asserted-by":"crossref","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","author":"S Hochreiter","year":"1997","journal-title":"Neural computation"}],"updated-by":[{"DOI":"10.1371\/journal.pcbi.1010050","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2022,4,21]],"date-time":"2022-04-21T00:00:00Z","timestamp":1650499200000}}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1010050","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,4,21]],"date-time":"2022-04-21T13:43:24Z","timestamp":1650548604000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1010050"}},"subtitle":[],"editor":[{"given":"Zhaolei","family":"Zhang","sequence":"first","affiliation":[]}],"short-title":[],"issued":{"date-parts":[[2022,4,11]]},"references-count":51,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2022,4,11]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1010050","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/2021.06.08.447505","asserted-by":"object"}]},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2022,4,11]]}}}