IBS及抑郁癥相關腸道菌群課件

1、IBS及抑郁癥相關腸道菌群的基因組學方法及若干研究進展Research focuses of my lab Genome and protein sequence analysis and proteinstructure analysis Bioinformatics methods of metagenomic analysisassociated with human health and environments Clinical data analysis and medicine informaticsIntestinal Microbiota- “Forgotten” Human

2、 Organ Gut microbes: 1014 cells cellnumbers10 times larger thanhuman; encode gene number 150times larger than humangenome(Baquero,2012). Influence alterations in host energybalance and immunity(Sanz,2014). Mysterious:80% gut microbes cantbe cultured in lab (Kellenberger,2001).1. Baquero F, Nombela C

3、. Clinical Microbiology and Infection, 2012, 18(s4): 2-4.2. Sanz Y, Olivares M, Moya-Prez , et al. Pediatric research, 2014.3. E. Kellenberger. EMBO reports. 2001, 2(1):57Hot Research Field 2008, NIH initiated HMP(Human Microbiome Project). 2008, European Commission initiated MetaHIT(Metagenomics of

4、 the Human Intestinal Tract consortium).Publication number related to intestinal microbiota (Sekirov,2009)1. Qin J, Li R, Raes J, et al. Nature, 2010, 464(7285): 59-65.2. Nelson K E et al. Science (New York, NY), 2010, 328(5981): 994.3. Gut Microbiota in Health and DiseaseGut Microbiota and Host Dis

5、ease Relationship: a). Dysbiosis triggers pathogenesis. b). Dysbiosis arises in parallel withpathogenesis. c). Disease causes shift in gut florastructure. d). Dysbiosis aggravates disease.Significance of Gut Flora Dysbiosis(Frank, 2011)1. Frank D N, Zhu W, Sartor R B, et al. Investigating the biolog

6、ical and clinical significance of human dysbiosesJ. Trends in microbiology, 2011, 19(9): 427-434.Gut Microbiota and Host DiseaseGut Microbes and Diabetes (Sanz ,2014)Gut Microbes and Obesity & Metabolic Syndrome (Devaraj,2013)1. Sanz Y, Olivares M, Moya-Prez , et al. Understanding the role of gut mi

7、crobiome in metabolic disease riskJ. Pediatric research, 2014.2. Devaraj S, Hemarajata P, Versalovic J. The human gut microbiome and body metabolism: implications for obesity and diabetesJ. Clinical chemistry, 2013, 59(4): 617-628.Metagenomics:Herein defined as the application of sequencing toDNA ob

8、tained from environmental or humanmicrobial samples.-Bypass the limitation of pure cultures sequencing-Short DNA reads by shotgun sequencing, especially byNGSSeveral metagenome sequencing projects in 2008 (Hugenholtz et.al. 2010)Metagenomics: method to analyze gutmicrobiotaSample CollectionDNA Filte

9、rDNA ExtractionPCRSequencingMetagenome Sequencing Progress (Wooley,2010)Metagenome Analysis Progress (Kunin,2008)1. Wooley J C, Godzik A, Friedberg I. A primer on metagenomicsJ. PLoS computational biology, 2010, 6(2): e1000667.2. Kunin V, Copeland A, Lapidus A, et al. A bioinformaticians guide to me

10、tagenomicsJ. Microbiology and Molecular Biology Reviews, 2008, 72(4): 557-578.Life scientists are starting to grapple with massive data sets,encountering challenges with handling, processing and movinginformation that were once the domain of astronomers and high-energy physicists.Nature 498 (13 June

11、 2013)The last week of April was designated Big Data Week. But inmodern biology, every week is big-data week.Nature 499 (4 July 2013)Short read data by NGS: bioinformatics challengesHiggins G., Human Genomes andBig Data Challenges, 2013,AssureRx Health Inc.Bioinformatics issues formetagenomics-Seque

12、ncing-Reads preprocessing-Short reads assembly-Gene prediction andannotation-Composition estimates-Binning /classification-Population analysis-Gene-centric analysis-Data compressionCurrent works in bioinformatics methods metagenomes-DNA short reads assembly methods: MAP and IntegMAP (Bioinformatics,

13、Zhu*, 2012; BMC Bioinformatics, Zhu*, 2015)-Gene prediction methods: MetaTISA (Bioinformatics, Zhu*, 2009),MetaGUN（BMC Bioinformatics, Zhu*, 2013）-Metagenomic sample comparison tool: MetaComp (2015)De Novo Assembly Methods for DNAShort Reads in MetagenomesSequence assembly plays an essential role in

14、 themetagenomics-Assemble short reads (25-1000 bp) into longer contigs (from 102bp to the whole chromosome) in order to provide more valuablegenomic content, which is essential for downstream analysis suchas gene finding and functional annotation.The information contained in different lengths of gen

15、omic DNAReference genome-based methods:-The comparative assembly approach such as AMOS uses areference genome or closely related species to align reads, wasapplied to facilitate assembly of short reads-The potential bias caused by phylogenetic complexity anddiversityDe novo methods:-The de novo asse

16、mbly methods are still regarded asirreplaceable tools for accurately assembling the novel genomicsequences that broadly exist in the metagenomic sequencing dataMAP (Metagenome Assembly Program) forSanger and 454 sequencing reads-A de novo assembly approach based on an improvedoverlap/layout/consensu

17、s (OLC) strategy incorporated withseveral special algorithms-Use the mate pair information, resulting in being moreapplicable to shotgun DNA reads currently widely used inmetagenome projects./MAP/Flowchart of MAPAssembly results of MAP on simulated Sanger reads (800 bp)Assembly results of MAP on sim

18、ulated 454 mate pair reads (200 bp)Results of extensive tests on simulated data showthat MAP can be superior to both Celera and Phrapfor typical longer reads by Sanger sequencing, aswell as has an evident advantage over Celera,Newbler and the newest Genovo, for typical shorterreads by 454 sequencing

19、.IntegMAP (Integrated MetagenomicAssembly Pipeline) for short reads by NGS-Developed a de novo pipeline, IntegMAP, for integratingindividual current assemblers that complemented the advantageseach in assembling metagenomic sequences-ABySS（Simpson et al., 2009）-CABOG (Miller et al., 2008)-IDBA-UD (Pe

20、ng et al., 2012)-MetaVelvet (Namiki et al., 2012)-SOAPdenovo (Li et al., 2010)Flowchart of IntegMAP High coverage ABySS IDBA-UD Low coverage IDBA-UD CABOGComparison of IntegMAP and other assemblies onsimulated metagenomic datasetTotalcoverCorr. N- Corr. N- E-sizelen at 10 len at 50 (bp)Num. of Total

21、Kbp / Identityerrors (%)coveredgeneserrorslength Mbp (bp) Mbp (bp)(Mbp)ABySS,k=31ABySS,k=61Bambus2CABOGIDBA-UDMetaVelvet,k=23MetaVelvet,k=61SOAPdenovo,k=23SOAPdenovo,k=61163.8 185,12285.5 222,5813,748 11,4664,192 15,3952,370 6,5315,713 10,1428,092 14,65142,37633,99740,139 259,32011,6546,71914.112.70

22、.998.642.155.999.899.999.599.899.799.899.999.899.9232.590,788244.8 139,195227.9 222,63147,96867,71323,97126,74714,25324,0812,4825,4163,271251 304.11,717 118.3182.85,4371,2749346898,62834576.3 121,245203.075.22,11689,8118796716,0781,92139.1Only contigs with length 200 bp are considered. “k=23”, “k=31

23、” and “k=61” in the first columndenote the assembler use the option of k-mer size at 23 bp, 31 bp and 61 bp. Bambus 2 uses unitigsfrom CABOG. Total cover length denotes the total length of reference sequences that are covered bycontigs. Corr. N-len denotes the corrected N-len size. E-size is also co

24、mputed using corrected contigs.Only complete covered genes are counted. Errors denote the structural errors in contigs. The errorrate is measured as the average distance between errors. Identity denotes the average identity of thealignments between contigs and references, where unmapped segments of

25、contigs are not considered.Values in bold indicate the best in the column.Assembly statistics and predicted gene number onhuman gut microbial metagenome dataset (SampleMH0012).Sum ofcontiglength(Mbp)158.7N-len at 5Mbp (bp)N-len at E-size (bp) Non-redundantNum. ofpredictedcompleteORFs50 Mbp(bp)ORFs p

26、redictedABySS, k=61aBambus2215,12550,90546,459177,46820,4698,90318,3666,427184,441336,604222,638339,336112,237226.6185.6277.2184,683119,907186,427CABOG12,73841,8316,964IDBA-UD23,970SOAPdenovoOnly contigs 500 bp are considered. Bambus 2 uses unitigs from CABOG. “k=51” denotes that237.434,5188,6795,16

27、6306,657135,644Met(aQVinelevteatl.u2s0e1s0o)bption of k-mer size 51 and “k=31” denotes that MetaVelvet uses option of k-mersize 31. The assembly generated by Qin et al (2010) is included, which is assembled byIntegMAP278.6242,60848,30339,156339,598186,997SOAPdenovo. In the column of non-redundant OR

28、Fs predicted, only ORFs 100 bp are counted.Last column lists the number of complete ORFs. The ORFs are predicted by MetaGeneMark (Zhu etal. 2010). Values in bold indicate the best in the column.aAssembly by ABySS was generated from the corrected reads from which many low coverage readsmay be exclude

29、d because we failed ran ABySS on the mixed reads.bAssembly by SOAPdenovo was directly downloaded from the publication of Qin et al. (2010).Taking advantage of the strength of each assemblerand the complementary among them, the IntegMAPpipeline improves largely in the metagenomicassembly performance

30、by improving assemblies onall sequencing depth levels.Compared with individual assemblers on bothsynthetic and real NGS metagenomic dataset,IntegMAP demonstrates its better performance ofgenerating assembly for both cover length andcontiguity with a high accuracy, in assembling NGSmetagenomic data.A

31、b initio Gene Prediction inMetagenomic DNA FragmentsAccurately identifying genes from metagenomicfragments is one of the most fundamental issues Most fragments are very short. Many sequences in metagenomicsequencing projects remain as unassembled reads or short-lengthcontigs. Therefore, lots of gene

32、s are incomplete with one or twoends exceed the fragments. Also, a single fragment usually containsonly one or two genes, non-supervised methods for single genomeswhich require many genes for model training are inapplicable forthis situation. The anonymous sequence problem, which means the sourcegen

33、omes of the fragments are always unknown or totally new,brings challenge on statistical model construction and featureselection.Evidence-based methodAb initio method-Evidence-based methods rely on homology searches includingcomparisons against known protein databases by BLASTpackages, CRITICA and Or

34、pheus.-Evidence-based methods can infer functionalities and metabolicpathways of the predicted genes via significant targets with ahigh specificity.-However, only the genes with previously known homologs canbe predicted by evidence-based methods , while the novel genes,which are very important to me

35、tagenomic studies, will beoverlooked.-Therefore, ab initio algorithms that can present much highersensitivity along with sufficient high specificity are indispensible.MetaGUN: gene prediction for metagenomicfragments based on SVM algorithmImplements by multi-strategy to predict genes:-Classifies inp

36、ut fragments into phylogenetic groups by a k-merbased sequence binning method.-Identifies protein coding sequences for each groupindependently with SVM classifiers that integrate entropy densityprofiles (EDP) of codon usage, translation initiation site (TIS)scores and open reading frame (ORF) length

37、 as input patterns.Then adjust TISs by employing MetaTISA.Flowchart of MetaGUNInput withmetagenomicdataTo identify protein-codingsequences, MetaGunbuilds the universalmodule and the novel genemodule. The former isbased on a set ofBinning based on k-RPS-BLAST formersdomainUniversalNovel genemodulemod

38、ulerepresentative species,while the latter is designedto find potentialORFindentificationfunctionary DNAsequences with conserveddomains.TIS relocating byMetaTISAOutputMetaGUNs performance on simulated metagenomic data Simulated metagenomic short-gun sequences Simulated fragments from 50 prokaryotic

39、genomes 4 kinds of read-length Sn=TP/(TP+FN), Sp=TP/(TP+FP), Hm=2SnSp/(Sn+Sp) For longer fragments, MetaGUN has better performancethan all other toolsApplication to human gut microbiome samples Two samples of human gut microbiome from two healthy humans (Gillet.al. 2006 Science) Potential novel gene

40、s: A：Genes with e10-5 searched in CDD database B：Genes annotated by IMG/M C：Genes with e10-5searched in NCBI NR database Potential novel genes：A-B-CSupporting findings for predicted novel genes-infB: corresponds to translation initiation factor IF-2, which isdifferent from the similar proteins in th

41、e Archaea and Eukaryotesand acts in delivering the initiator tRNA to the ribosome-PRK12678: corresponds to transcriptional terminator factor Rho;-Several domains from DNA polymerase like PRK05182,PRK12323./MetaGUN/MetaTISA/Brain-gut axis disorder andIntestinal microbiologyZhu LabDuan LabIrritable bo

42、wel syndrome(IBS)AbnormalMotilityHigh PrevalenceHigh VisceralSensitivityIntestinalInflammationPsychologicalFactorsLow Cure Rate1. Mayer, Emeran A., Tor Savidge, and Robert J. Shulman. Braingut microbiome interactions and functional bowel disorders. Gastroenterology146.6 (2014):1500-1512.IBS and Gut

43、Microbiota MicroecologyThe imbalance of intestinal micro-ecology in IBS patients.The ratio of Firmicutes/Bacteroidetes is significantly changed.1. Jeffery, Ian B., et al. An irritable bowel syndrome subtype defined by species-specific alterations in faecal microbiota. Gut 61.7 (2012): 997-1006.2. Ca

44、rroll, Ian M., et al. Alterations in composition and diversity of the intestinal microbiota in patients with diarrheapredominant irritable bowelsyndrome. Neurogastroenterology & Motility 24.6 (2012): 521-e248.3. RajiliStojanovi, Mirjana, et al. Global and deep molecular analysis of microbiota signat

45、ures in fecal samples from patients with irritable bowelsyndrome. Gastroenterology 141.5 (2011): 1792-1801.Mental Disorder Accompanied with IBSIllness TypeDetection Rate（%）（N=246）FDIBS44.639.8FD+IBSCIDI-3.0 (MD)34.721.845.139.2Somatoform DisorderComorbidity Rate of IBS and Mental Disorder*Abbreviati

46、on: FD- Functional Diarrhea PD- Personality Disorder MD- Mental Disorder*This survey is conducted by Department of Gastroenterology, Peking University Third HospitalDepression and Gut Microbiota MicroecologyThe imbalance of intestinal micro-ecology in Depression patients.The ratio of Bacteroidetes/F

47、irmicutes is significantly changed.1. Finegold, Sydney M., et al. Pyrosequencing study of fecal microflora of autistic and control children. Anaerobe 16.4 (2010): 444-453.Brain-Gut Axis DisorderABrain-Gut Axis :Bidirectional Interactions between Brain and Gut (Mayer,2014)1. Mayer, Emeran A., Tor Sav

48、idge, and Robert J. Shulman. Braingut microbiome interactions and functional bowel disorders. Gastroenterology146.6 (2014):1500-1512.Research Goal Compare the microbial community structure in IBS,Depression and Comorbidity patients gut. Clarify the gut microbes signature of patients. Identify pathog

49、enic bacteria. Analyze correlation between clinical symptoms and gut microbes. Explore new target for clinic treatment. Compare the functional difference of gut flora in IBS,Depression and Comorbidity patients gut. Clarify the correlation between gut flora metabolic function and hostdisease severity

50、. Analyze causality of gut microbiota and disease. Explore new therapies targeted on regulating gut microbiota metabolicfunctions. Determine the importance of gut flora in the morbidity ofbrain-gut axis disorder related diseases.Gut FloraBrain-gut axis disorder andIntestinal microbiologyMicrobial st

51、ructural dysbiosisMicrobial functional disorderPatients Symptoms DescriptionMicrobial Structure VariationPhylum Distribution (*: P0.05)ARarefraction CurveBDSpecies AbundanceCFamily DistributionGroup Genus AbundanceHeatmap Analysis on Genus Abundance of 100 SamplesBacterial Taxa differ between IBS-D,

52、Depression, COMO and Health ControlMean taxa numberPhylumGenusControlIBSDepressionCOMOPrevotellaBacteroidesParaprevotella170.421861.218.321284.05a3442.67a34.3b1142.13a4800.33a39b970.4a4318.04a29.4bBacteroidetesBlautiaLachnospiracea_incertae_sedisClostridium IV117.26594.2121.3213.7911.37731.631.4724.

53、3a186.9a19.324.6b3.2b346.7b1.6427.89a427.660.89a15.4613.65240.67b4.22b216.92a400.2821.1119.6416.22FirmicutesClostridium XlVaClostridium XVIIIFaecalibacterium865.215.08bFlavonifractorTM7_genera_incertae_sedisGemmiger1.713.531.5812.790.11a2.33b2.52b14.83Proteobacteriaa:P0.01 compared with controlb:P0.

54、05 compared with controlSample Cluster By Microbial Composition100 samples are grouped into 3 clusters distinguished by the genus dominanceSample Distribution in 3 ClustersMost IBS and Depression patients are in Cluster I and II, while health controls are in Cluster III.Sample Difference among Clust

55、ersA Distance among ClustersB Species Abundance among ClustersC Bacteroides Abundance in Cluster I, IID Prevotella Abundance in Cluster I, IIPhylogenetic Differences betweenPatients and Health ControlsOver Expression of MCP-1 & MIP-a associatedwith Cluster and Clinical SymptomsPathogenic Relevance A

56、nalysisIBS-DCluster I Cluster IIP0.05DepressionCOMOGroup I+IIPhylumGenusCluster ICluster IIP0.05P0.05BacteroidesBarnesiella0.83-0.580.85-0.51-0.38BacteroidetesOdoribacter-0.39Prevotella-0.93-0.95-Tannerella-0.54Blautia-0.49-0.44-Clostridium XlVaCoprococcusDorea-0.48-0.39-0.31-FaecalibacteriumFlavoni

57、fractorLachnospiracea_incertae_sedisOscillibacter-0.54-0.44-0.440.42-_-Firmicutes-0.52-0.36-0.47-0.38-Roseburia-0.32-StreptococcusTuricibacter-0.380.67-0.37Veillonella-0.57-LentisphaeraeProteobacteriaTM7VictivallisCampylobacterGemmiger-0.390.54-0.44-0.37-0.40-0.40-0.57-HaemophilusTM7_genera_incertae

58、_sedis-Pearson correlation coefficient was used to measure the correlation between microbialgenera and different clusters in IBS-D, depression and COMOMicrobe Interactions in IBS and HealthBrain-gut axis disorder andIntestinal microbiologyMicrobial structural dysbiosisMicrobial functional disorderFu

59、nctional Distribution of Microbial ContigsCellular ProcessingEnvironmental InformationProcessingGenetic InformationProcessingHuman DiseasesMetabolismOrganismal SystemsFunctional Distribution Variation among PatientsGroupsCOMO CureIBSCureCOMO OrigDepressionHealthIBS OrigPotential pathogenicity of gut microbes*Significant Higher Abundance of Genes in Patients Gut Flora*:P0.05FhuE : Responsible for transport of ferric coprogen andferric-rhodotorulic acid. And iron ion are essentialfor reproduction