Multi-Stage Probabilistic Bipartite Graph Algorithm - Effect of Herbal Medicines on the Gut Ecosystem
Total Page:16
File Type:pdf, Size:1020Kb
Multi-Stage Probabilistic Bipartite Graph Algorithm - Effect of Herbal Medicines on the Gut Ecosystem Presenter: Suganya Chandrababu 03/06/2020 University of Nebraska at Omaha GRACA 2019 Straightforward Overview n Purpose/value of the study n Effect of Herbs on the Gut Ecosystem n Our Contribution n Multi-stage Graph Probabilistic Algorithm n Dataset Used n Results and Discussion n Conclusion n Limitations n Future Work 03/06/2020 University of Nebraska at Omaha GRACA 2019 Why do we care? n Such imbalance could result in intestinal and extra-intestinal disorders, including inflammatory bowel disease (IBD), diabetes mellitus (DM), obesity, etc. 03/06/2020 University of Nebraska at Omaha GRACA 2019 Pharmacological Effects of Natural Products 03/06/2020 University of Nebraska at Omaha GRACA 2019 Herbs Modulate Gut Bacterial Metabolism 03/06/2020 University of Nebraska at Omaha GRACA 2019 Bibliographic DB Limitation Structure DB MEDLINE, Microarray DB Interoperability PubMed NDB, GEO, PDB Data HUGE Unstructured Metabolic DB Enzyme DB Applicability KEGG, MetaCyc BRENDA, Scope of data LimitationsMPBA ExPasy coverage Inconsistent Genome DB Ontology SWISS Genome DB Prot, Data EMBL GIB, NCBI Organization OMIM PubChem Chemical DB Disease DB 03/06/2020 University of Nebraska at Omaha GRACA 2019 Why Hear the Talk? To n understand a comprehensive method that integrates knowledge from multiple sources. n know how to characterize the microbial dysbiosis. n evaluate the pharmacological value of herbal medicine. n learn how and which herbs can modulate the gut microbial functions. 03/06/2020 University of Nebraska at Omaha GRACA 2019 What is our contribution? We n provide a systems-level characterization of the effects of the small molecules in herbs on the bacterial genes and metabolic pathways. n propose a computational network-based approach- the Multi-stage Probabilistic Bipartite graph Algorithm (MPBA), which integrates the complex heterogeneous metabolism/pathway data from herbs to gut bacteria. 03/06/2020 University of Nebraska at Omaha GRACA 2019 Background Work u The human colon modulates of gut immune function, protects the host against pathogens and diseases. [D. C. Savage, et al., 1997, R. E. Ley, et al., 2010] u Variations in products of bacterial metabolism have been implicated as a causative agent in serious diseases, including cancer (example: Helicobacter pylori). [Flint, Harry J., et al., 2012] u Herbal medicine are useful in the treatment of dysbiosis and serve as a multi-targeting complementary/alternative medicine/probiotic in the management of colon health, [M. Sałaga, et al., 2014, F. Ke, et al., 2012] 03/06/2020 University of Nebraska at Omaha GRACA 2019 Multi-Stage Probabilistic Bipartite Graph Algorithm 03/06/2020 University of Nebraska at Omaha GRACA 2019 Our Prediction Model Bacterial Bacteria Enzymes Pathways metabolism MPBA Bacteria Pathways Herbs Compounds Targets Herbal MPBA pharmacology Herbs Targets Pathways MPBA Herbs Pathways Herbs modulating bacterial functions Herbs Pathways Bacteria 03/06/2020 University of Nebraska at Omaha GRACA 2019 The MPBA Pathways B1 P1 B1 H1 H1 B2 Bacteria Graph Approximation B2 Bacteria Herbs Herbs Hi Pq Hi Bj Bj , where Pk Hm Bn Hm Bn iq x yjq zij (a) (b) It denotes the degree of the pathway nodes. Matrix Form , where ∑ is a diagonal matrix , with 03/06/2020 University of Nebraska at Omaha GRACA 2019 The Divergence Loss Function u To evaluate how well the algorithm models the data, we use the divergence loss function u The approximation , can be done minimizing Divergence Loss , which equal zero when 03/06/2020 University of Nebraska at Omaha GRACA 2019 Probabilistic Graph Model u We can interpret the same equation using Markov random walks on graphs as done using PLSA [T. Hoffman, 200]. u The weight zij is essentially a quantity proportional to the stationary probability of direct transitions between hi and bj, denoted by p(hi , bj). zij = p(hi , bj) u In a tripartite graph there is no direct edge between the hi node and the bj node, and they can only be connected through the pathway nodes pq 03/06/2020 University of Nebraska at Omaha GRACA 2019 Probabilistic Graph Model Transition probability Tripartite Network Back Substitution 03/06/2020 University of Nebraska at Omaha GRACA 2019 Model Fitting Using EM Algorithm 03/06/2020 University of Nebraska at Omaha GRACA 2019 Bayesian Formula u For the E-step we apply Bayes’ formula to obtain u And in the M-step one has to maximize the expected complete data log-likelihood given by, u where is the frequency or the number of paths from herb to bacteria through the pathway node. 03/06/2020 University of Nebraska at Omaha GRACA 2019 Dataset Used 03/06/2020 University of Nebraska at Omaha GRACA 2019 Differential Abundant Pathways in IBD Nitrogen metabolism Fructose and mannose metabolism Riboflavin metabolism Sulfur metabolism Bacterial secretion system Cysteine and methionine metabolism Butanoate metabolism UC CD without illeal Lysine biosynthesis CD with illeal Healthy Pentose phosphate pathway 2.X. C. Morgan, T. L. Tickle, H. Sokol, D. Gevers, K. L. Devaney, 3.D. V. Ward, J. A. Reyes, S. A. Shah, N. LeLeiko, S. B. Snapper, et al., “Dysfunction of the intestinal microbiome in inflammatory bowel disease and treatment,” Genome biology, vol. 13, no. 9, p. R79, 2012. 03/06/2020 University of Nebraska at Omaha GRACA 2019 Dataset and Source Herbal Bacteria compounds target harbors 273 262 bacterial enzymes enzymes belonging to belonging to the 8 Most the pathways 235 pathways 24 Culinary affected bacterial herbs, 132 Pathways species bioactive and and 273 compounds enzymes bacterial and 262 in IDB enzymes compound Provides Targets target enzymes Source: Source: KEGG Uniprot, Source: NCBI PhytoChem Genome NAL, PubChem 03/06/2020 University of Nebraska at Omaha GRACA 2019 Results and Discussion 03/06/2020 University of Nebraska at Omaha GRACA 2019 Case Study n Most Affected Microbial Metabolic Pathways and Enzymes In Both UC And CD n Routes were traced through the metabolic model of the gut microbiota, and the organism(s) harboring the necessary genes identified. 03/06/2020 University of Nebraska at Omaha GRACA 2019 Bacterial metabolism Bacteria Enzymes Pathways Veillonella dispar ATCC 17748 Lactobacillus delbrueckii subsp. bulgaricus Veillonella sp. oral taxon 780 str. F0422 Streptococcus vestibularis F0396 PB2003/044-T3-4 Clostridium sp. SS2/1 Peptoniphilus lacrimalis 315-B Lactobacillus crispatus 125-2-CHN 1 Dialister microaerophilus1 UPII 345-E Streptococcus sp. 2_1_36FAA 1 Coprococcus eutactus1 ATCC 27759 Lactobacillus parafarraginis F0439 1 Finegoldia magna ACS-171-V-Col3 1 EC 2.5.1.78 Eubacterium saphenum ATCC 49989 LactobacillusErysipelotrichaceae iners DSM 13335 bacterium 3_1_53 1 Anaerococcus tetradius ATCC 35098 1 1 1 1 1 1 u The network 1 1 Lactobacillus gasseri 224-1 EC 4.3.1.17 Faecalibacterium cf. prausnitzii KLE1255 EC 2.5.1.9 EC 2.1.1.10 Streptococcus sanguinis 1SK115Veillonella sp. oral taxon 158 str. F0412 Veillonella parvula ATCC 17745 1 Peptoniphilus harei ACS-146-V-Sch2b1 1 Listeria grayi DSM 20601 EC 3.2.2.9Streptococcus mitis SK1073 1 Coprobacillus sp. 8_2_54BFAA EC 3.5.4.25 1 1 map00740 1 1 1 EC 2.1.1.141 1 1 consists of 8 Streptococcus sanguinis1 VMC66 1 Selenomonas artemidis F0399 Roseburia inulinivorans DSM 16841 1 EC 2.7.7.21 1 1 1 1 Lachnospiraceae bacterium oral taxonEC 082 2.7.1.26 1 1 1 EC 2.1.1.37 1 str. F0431 Lactobacillus iners LactinV 01V1-a EC 5.3.1.23 Streptococcus peroris ATCC 700780 1 1 1 EC 2.5.1.16Streptococcus sanguinis1 SK408 1 1 1 EC 4.1.99.12 Pseudoramibacter alactolyticus ATCC 23263 1 EC 2.5.1.6 Streptococcus parasanguinis SK236 1 1 1 1 Bacillus 1sp. 2_A_57_CT2 1 1 1 pathways, 273 Leuconostoc mesenteroidesClostridium subsp. sp. 7_3_54FAA EC 2.3.1.31 EC 1.2.1.11 1 Veillonella sp. 6_1_27 1 Streptococcus anginosus F0211 EC 1.13.11.53cremoris1 ATCC 19254 Streptococcus infantis ATCC 700779 1 1 1 1 11 1 Streptococcus1 vestibularis ATCC 49124 1 1 1 Oribacterium sp. oral taxon 108 str. F0425 1 1 1 Lactobacillus1 kisonensis F0435EC 1.1.1.193 1 EC 1.17.1.8 1 EC 1.1.1.3 1 1 EC 3.5.4.31 1 Lactobacillus ultunensis1 DSM 16047 1 map00270 1 1 1Paenibacillus sp. oral taxon 786 str. D14 Lactobacillus amylolyticus DSM 11664 1 1 bacterial Streptococcus sanguinis1 SK72 1 1 1 1 1 1 1 1 Streptococcus sanguinis SK1087EC 2.5.1.48 Streptococcus oralis SK313 EC 6.3.2.3 1 EC 2.6.1.42 1 Coprococcus1 comes ATCC 27758 Paenibacillus sp. HGF5 1 1 EC1 2.6.1.83 map00300 1 1 1 1 Streptococcus cristatus ATCC 51100 Dorea longicatena DSM 138141 1 1 EC 2.7.2.4 Erysipelotrichaceae bacterium 6_1_45 Faecalibacterium prausnitzii A2-165 Lactobacillus antri DSM 16041 1 1 Eggerthia catenaformis OT 569 = DSM 1 1 EC 2.6.1.1 1 1 enzymes, 235 1 20559 1 1 1 1 1 1 1 Streptococcus sanguinis SK1056EC 2.3.1.89 EC 6.3.2.2 1 1 Coprobacillus1 sp. 3_3_56FAA Paenibacillus sp. HGF7 1 Fusobacterium periodonticum ATCC 336931 1 EC 2.3.1.30 Streptococcus oralis ATCC 35037 EC 4.4.1.21 1 1 EC 6.3.2.10 1 1 Streptococcus sanguinis SK330 1 1 1 1 1 1 Streptococcus infantis SK1076 Subdoligranulum variabile DSM 15176 Enterobacteriaceae bacterium 9_2_54FAA 1 1 1 1 1 1 1 1 1 Centipeda periodontii DSMEC 5.3.1.92778 bacterial species, 1 1 Granulicatella adiacens ATCC 49175 1 EC 4.1.1.5 Anaerococcus1 prevotii ACS-065-V-Col13 EC 4.1.1.20 1 EC 5.4.2.7Bacillus smithii 7_3_47FAA 1 Streptococcus sanguinis SK1057 Streptococcus equinus ATCC 9812 1 EC 2.5.1.491 map00920 1 1 1 1 1 Anaerococcus hydrogenalis Leptotrichia hofstadii F0254 1 EC 1.1.1.27 ACS-025-V-Sch4 1 EC 1.1.99.3 1 EC 2.5.1.47 1 EC 4.2.1.17 and 18060 Lachnospiraceae bacterium 2_1_46FAA 1 Clostridium celatum DSM 1785 1 Selenomonas sp.