High throughput biological sequence collection, analysis and its application in analysis of oral microbiota

Transcription

1 High throughput biological sequence collection, analysis and its application in analysis of oral microbiota Chaochun Wei ( 韦朝春 ) Department of Bioinformatics and Biostatistics Shanghai Jiao Tong University Fall 2013

2 Contents Background A brief history of genomics technology developments Introduction to metagenomics Oral metagenomics Summary

3 The ultimate goal is for sequencing to become so simple and inexpensive that it can be routinely deployed as a general-purpose tool throughout biomedicine., Research applications will include characterizing genomes, epigenomes and transcriptomes of humans and other species, as well as using sequencing as a proxy to probe diverse molecular interactions. Eric S. Lander, 2011, Initial impact of the sequencing of the human genome Nature

4 Genomics Technology developments

5 Milestone of Genomics Technology Affy launches Gene Expression microarrays First microarray publication - on Arabidopsis Affy & ILMN both launched 100K genotyping arrays The Sequencing Shake up!! ABI commercializes first automated DNA sequencer Hapmap project launched ILMN launches gene expression arrays ILMN bought Solexa; launches GA Roche GS FLX launched ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

6 Milestone of Genomics Technology Affy launches Gene Expression microarrays First microarray publication - on Arabidopsis Affy & ILMN both launched 100K genotyping arrays The Sequencing Shake up!! ABI commercializes first automated DNA sequencer Hapmap project launched ILMN launches gene expression arrays ILMN bought Solexa; launches GA Roche GS FLX launched ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

7 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays First microarray publication - on Arabidopsis 1986, ABI created the first automated DNA sequencer Hapmap project launched ILMN launches gene expression arrays Affy & ILMN both launched 100K genotyping arrays ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

8 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays First microarray publication - on Arabidopsis Hapmap project launched Affy & ILMN both launched 100K genotyping arrays In 1994, NCBI created the national DNA ILMN launches gene expression database called Genbank arrays ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

9 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays First microarray publication - on Arabidopsis Hapmap project launched Affy & ILMN both launched 100K genotyping arrays In 1998, ABI 3700 DNA sequencer was launched tothe market. HGP became a large scale. The intense competition between the ILMN launches science community and the industry gene expression arrays accelerated the HGP greatly. ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

10 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays HGP and First microarray a private company publication - on Arabidopsis Celera Genomics published the draft of HG simultaneously. China finished 1% of the HG. Hapmap project launched ILMN launches gene expression arrays Affy & ILMN both launched 100K genotyping arrays ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

11 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays First microarray In 2002, Human publication - on haplotype Arabidopsis project started. It finished in China finished 10%. Hapmap project launched ILMN launches gene expression arrays Affy & ILMN both launched 100K genotyping arrays ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

12 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays First microarray In 2006, Illumina publication - on Arabidopsis launched GA, the first NGS sequencer. Hapmap project launched ILMN launches gene expression arrays Affy & ILMN both launched 100K genotyping arrays ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

13 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays In First 2007, microarray Roche GS publication - on Arabidopsis FLX(454) and ABI Solid 1.0 were launched to market. Hapmap project launched ILMN launches gene expression arrays Affy & ILMN both launched 100K genotyping arrays ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

14 Milestone of Genomics Technology ABI commercializes first automated DNA sequencer Affy launches Gene Expression microarrays Hapmap project launched Affy & ILMN both launched 100K genotyping arrays The First microarray 3 rd generation of publication - on Arabidopsis sequencer may be launched to the ILMN market launches in gene expression 2014? arrays ILMN bought Solexa; launches GA Roche GS FLX launched The Sequencing Shake up!! ILMN HiSeq 2000 launched In the coming future Low hanging fruit: cystic fibrosis mutation identified Rise of Genbank databases from DNA sequencing 3700 DNA Analyzer in Human Genome Project; DNA sequencing goes industrial Human Genome Project & Celera Genomics completes first draft genome Hapmap 1 st phase data release ABI SOLiD 1.0 Launched! Rise of Genome Wide Association Studies (GWAS) SOLiD 3.0: 100GB out of the box! The 3 rd Generation Sequencing will be launched

15 Cost of per base and per human genome dramatically dropped Innovation of NGS throughput Cost of per Human Genome Throughput (Gb) Gb-300Gb $M 100, , years ~$3,000,000, , Moore s Law <2 weeks ~$1, Gb Gb 6Gb he deep sequencing technology become more and more popular in translational medicine research because of its lower price and high-throughput

16 The revolution caused by Next- Generation Sequencing (NGS) < <? NGS: 2 nd generation One staff, one machine 3 rd generation: Oxford Nanopore , Sanger sequencing requires industrialized lab and many staffs NATURE METHODS,16 VOL.5 NO.1 JANUARY

17 NGS platforms 3 rd generation: Oxford Nanopore MinION Applied Biosystems ABI 3730XL Roche / 454 Genome Sequencer FLX HeliScope Single Molecule Sequencer Illumina / Solexa Genetic Analyzer Applied Biosystems SOLiD 17

18 Comparison of NGS and traditional sequencing platform Platform Sanger 454 Solexa SOLiD Read length (bp) # of reads/run ,000 2,000,000, ,000,000,0 00 Error rate 10^-3 <10^-2 ~10^-2 ~10^-2 Cost($/Mbp) 5000 ~5 ~0.6 ~0.2 Time/run ~3 h ~7 hours 2-14 days 3-14 days Throughput 100Kb ~1Gb ~600Gb Gb

19 Metagenomics ( 元 / 基因组学 )

20 Microbes in a natural environment in community Complex system Structure and function diversity Organism composition Gene composition Function composition <1% can be isolated and cultured 20

21 Metagenome and Metagenomics Metagenome All genetic materials in an environmental sample Metagenomics Direct sequence and analyze metagenomic sequences Can study the composition structure of microbial communities in natural environments Thousands of new organisms have been found Changing our view about the world of life Provide new viewpoints and methods for environment, energy, and health related research areas 21

22 Typical Sources of Metagenomes Soil samples Sea water samples Seabed samples Air samples Medical samples Ancient bones Human microbiome

23 Human contains not only the human genome In a healthy adult Microbial cells are 10 times more than human cells Most of them are in gut 23

24 Three basic questions in Metagenomics Who is in there? Metagenome binning What are they doing? Function annotation How do they compare? Comparative metagenomics

25 Metagenome sequence analysis Sequencing data quality control Sequencing simulation Finding community structure Metagenome sequence classification Finding the functional composition structure How do they compare

26 Raw metagenome data Input: Sponge metagenome sequence dataset 164,421 reads (23.2M) Sequence length: Input file in FASTA format (our pipeline accepts FASTQ file too)

27 Sequencing quality control Before quality control: Low quality on both ends of reads (top) A large number of reads have a hyper low average quality score (bottom) FastQC: fast quality control for sequence data

28 Sequence quality control After quality control: Low quality bases eliminated (top) Reads with a average quality score less than 20 have been removed (bottom) FastQC: fast quality control for sequence data

29 Metagenome sequence analysis Sequencing data quality control Metagenome sequencing simulation Finding community structure Metagenome sequence classification Finding the functional composition structure How do they compare

30 NGS sequencing bias Figure 2.The distribution of quality values at each base. X axis: the coordinates of reads (0-based); Y axis: the PHRED scores. The blue dots represent the average quality values and the red dots represent the median. (A) Illumina; and (B) 454 sequencing platform. Jia B, Xuan L, Cai K, Hu Z, Ma L, et al. (2013) NeSSM: A Next-Generation Sequencing Simulator for Metagenomics. PLoS ONE 8(10): e75448.

31 Metagenome sequencing simulation The comparison of sequencing coverage before and after simulation. X axis: the coordinate of the genome of Acinetobacter baumannii ATCC Each interval contains 100 bases and only the first 3,000 intervals are shown; Y axis: the read numbers mapped in each interval. A: the sequencing coverage in the Dataset F; B: NeSSM; C: MetaSim ; D: GemSIM ; E: Grinder; and F: pirs. Jia B, Xuan L, Cai K, Hu Z, Ma L, et al. (2013) NeSSM: A Next-Generation Sequencing Simulator for Metagenomics. PLoS ONE 8(10): e75448.

32 The distributions of read lengths. X axis: the lengths of reads. Each interval is 10 bps; Y axis: the number of reads with lengths in a certain interval. Jia B, Xuan L, Cai K, Hu Z, Ma L, et al. (2013) NeSSM: A Next-Generation Sequencing Simulator for Metagenomics. PLoS ONE 8(10): e75448.

33 Software Platform Read number Read length Time(s) NeSSM_CPU Illumina 90 million NeSSM_GPU Illumina 90 million MetaSim Illumina 90 million 36 3,821 GemSIM* Illumina 90 million 36 90,600* Grinder* Illumina 90 million 36 2,143,078* NeSSM_CPU million 250 5,560 NeSSM_GPU million MetaSim million ,968 GemSIM* million ,359* Grinder* million 250 2,236,412* Comparison of the speed of NeSSM and existing tools on HC metagenome simulation.*: predicted by a linear extension of the times for a series of small datasets. Jia B, Xuan L, Cai K, Hu Z, Ma L, et al. (2013) NeSSM: A Next-Generation Sequencing Simulator for Metagenomics. PLoS ONE 8(10): e75448.

34 Metagenome sequence analysis Sequencing data quality control Metagenome sequencing simulation Finding community structure Metagenome sequence classification How do they compare

35 Q1:WHO IS IN THERE?

36 Two types of metagenomics methods 1. Marker gene sequencing: 16S rrna, 18S rrna, 2. Whole genome shotgun sequencing

37 Variant regions of 16S rrna gene We find that different taxonomic assignment methods vary radically in their ability to recapture the taxonomic information in full-length 16S rrna sequences: most methods are sensitive to the region of the 16S rrna gene that is targeted for sequencing, but many combinations of methods and rrna regions produce consistent and accurate results. To process large datasets of partial 16S rrna sequences obtained from surveys of various microbial communities, including those from human body habitats, we recommend the use of Greengenes or RDP classifier with fragments of at least 250 bases, starting from one of the primers R357, R534, R798, F343 or F517. Accurate taxonomy assignments from 16S rrna sequences produced by highly parallel pyrosequencers. Liu Z, DeSantis TZ, Andersen GL, Knight R. Nucleic Acids Res Oct;36(18):e120. Epub 2008 Aug 22.

38 Limitations of 16S classification The copy number of 16S rrna gene can vary by an order of magnitude between bacterial species PCR-induced biases.

39 Environmental Shotgun Sequencing Sampling from habitat; Filtering particles, typically by size; DNA extraction and lysis; Cloning and library; Sequence the clones; Computational analysis A primer on metagenomics. Wooley JC, Godzik A, Friedberg I. PLoS Comput Biol Feb 26;6(2):e Review.

40 Metagenome sequence analysis Sequencing data quality control Finding community structure Metagenome sequence classification Finding the functional composition structure How do they compare

41 Background Metagenomic sequence classification Assign sequences to groups representing the same or similar taxa A prerequisite for genome assembly and biological diversity finding for an environment Two types of methods 1. Alignment based 2. Sequence composition feature based

42 Methods based on alignment ~5% <1%

43 Methods based on alignment Accuracy not accurate for those w/o similar genomes available Computation complexity? BLAST: too slow Bowtie: fast, but can t align those with indels or many mismatches

44 Sequence composition based methods Advantages: Alignment free Can deal with sequences from unknown genomes Existing methods Phymm (Brady A, Salzberg SL, 2009, 2011, Nat Methods) Computation complexity? Fast, but still takes weeks for Phymm to analyze one run of 454 sequencing

45 Graphics Processing Unit (GPU) GPU devotes more transistors to computing Many cores(a few hundred to a few thousand cores) Control Cache ALU ALU ALU ALU DRAM CPU Nvdia Tesla 2050 DRAM 448 cores, 3GB memory GPU

46 GPU computing power Floating-Point Operations per Second of the CPU and GPU

47 GPUs for Bioinformatics Applications 3D display acceleration Parallel computing (especially for little I/O, and big computing) Bioinformatics applications (speedup) Molecular structure computing, molecular dynamic simulation (~200) Sequence alignment(hammer, ~100) Database search (?) High-throughput sequencing data analysis (~20) Metagenomic sequence classification(~20)

48 Goal: classify short sequences Databases kmms for genomes NCBI Taxonomy table Input: short sequences Output: taxonomy of each sequence Available at: b/software/metabing/metabing.php MetaBinG: Using GPUs to accelerate metagenomic sequence classification

49 MetaBinG: Method P( X Kmer ) i j k F ( Kmer, X ) i k j F ( Kmer ) i k

50 MetaBinG: Method l k 1 S ln( p ( X Kmer )) i i k j j 0

51 Training and test datasets 1212 bacterial genomes 390 genomes were removed 468 for training, 354 for test Simulation dataset 6640 test reads for each of the ten different lengths Compare with Phymm *A. Brady and S. L. Salzberg. Phymm and PhymmBL. Nature Methods Vol. 6, No. 9, pp (September 2009)

52 Result comparison Phymm * MetaBinG Sequence Length Accuracy Time Accuracy Time (bps) (%) (s) (%) (s) Speedup Comparison of Phymm and MetaBinG with accuracy at phylum level Jia, P., Xuan, L., Liu, L., Wei, C.*, 2011 PLoS ONE, 6(11): e25353 * Brady A, Salzberg SL, 2009, Nat Methods, 6: * Brady A, Salzberg S 2011, Nat Methods, 8: 367.

53 MetaBinG for a real dataset Biogas reactor dataset (Schluter, et al., 2008) 616, reads Average length 230 bps All 1212 genomes are used for classification Computing time: MetaBinG: 248 seconds Phymm*: 4days 5 hours and 56 seconds Speedup: ~=1500 Jia, P., Xuan, L., Liu, L., Wei, C.*, 2011 PLoS ONE, 6(11): e25353 * Brady A, Salzberg SL, 2009, Nat Methods, 6: * Brady A, Salzberg SL, 2011, Nat Methods, 8: 367.

54 MetaBinG for a real dataset 14 of the top 15 phyla generated by Phymm was in the list of top 15 produced by MetaBinG. The relative ranks for these phyla varies at most by a value of two.

55 Contribution of GPUs MetaBinG versions Single threaded CPU version Parallel CPU version** GPU version Speed up 1 ~24 ~600 The contribution of GPUs: ~25 times speedup* The tests were on the biogas dataset. The real values may various. Parallel CPU version was implemented with BLAS library from MKL Intel.

56 MetaBinG for 60GB sequencing data Dataset: 0.48 billion of Solexa Illumina reads 100 bps in average Total size: 60GB Computing time: Phymm: 5 CPU years (estimated) MetaBinG: ~30 hours

57 Summary MetaBinG s accuracy is close to Phymm; MetaBinG is at least 2 orders faster than Phymm GPUs can speed up ~25 times for metagenome sequence classification GPUs can be applied to more bioinformatics areas Sequence alignment

58 Q2:WHAT ARE THEY DOING?

59 Metagenome sequence analysis Sequencing data quality control Metagenome sequencing simulation Finding community structure Metagenome sequence classification Finding the functional composition structure How do they compare

60 Gene calling Homology search, BLAST Sensitivity is low Specificity is high Ab initio gene prediction, Markov models.

61 Gene prediction methods used in metagenomic projects

62 Gene prediction on unassembled single reads and assembled contigs

63 Blast Megan MG-RAST OR Skip the gene calling step.

64

65 Method ipath Compare pathways of different parts Only A, here is Bacteria Only B, here is Pathogen A and B both (Since we use a small random dataset here, only aligned bacteria sequences are enough for downstream analysis, e.g. KEGG ortholog searching, while pathogen DB includes all bacteria data too. So here A & B actually are the same thing.)

66 Method - ipath Output ipath input file (by ipath.pl) KEGG Pathway ID Color Width = 2*log(amount) Opacity between 0-1

67 Bacteria: pathway Results KEGG pathway

68 Interactions between a host and its environment microbiota Red is for host only, green is for environment microbiota only, and blue is for both. The thickness of the lines is for the relative abundance.

69 Metagenome sequence analysis Sequencing data quality control Metagenome sequencing simulation Finding community structure Metagenome sequence classification Finding the functional composition structure How do they compare

70 Q3:How Do They Compare?

71 MEGAN MG-RAST JGI IMG CAMERA Tools for Metagenomics

72 Metagenomics and pathogen genomics Metagenomic data collection and analysis system Meta-All: Using all bacterial genomes available Community composition structure analysis Metagenome sequence classification Metagenome sequence assembly Functional element finding in Metagenome Unknown pathogen identification Metagenomics-based pathogen identification 72

73 16S-rRNAs in bacterial genomes 73

74 Meta-All vs. Megan* *Mitra S, Klar B, Huson DH (2009), Bioinformatics 25:

75 Metagenomics for oral microbiota

76 Materials and Methods Samples 60 children with ages from 3 to 6 34 boys, 26 girls MN (boy, no caries, n=17), MC (boy, with caries, n=17) FN (girl, no caries, n=11) FC (girl, with caries, n=15) Sample collection and preparation Saliva and plaque Sequencing 186,787 reads of V3 regions of the 16S rrna gene

77 The diversity of oral microbiota Ling et al., Microbial Ecology, 2010, 60(3): Streptococcus Prevotella Neisseria Pasteurella Veillonella Porphyromonas Leptotrichia Rothia Capnocytophaga Granulicatella Fusobacterium Thiomonas Gemella Actinomyces Hallella Corynebacterium Kingella Anaerosporobacter Campylobacter Planobacterium Oribacterium Tannerella Peptostreptococcus Cardiobacterium TM7_genera_incertae_sedis Atopobium SR1_genera_incertae_sedis Chryseobacterium Catonella Abiotrophia Solobacterium Dysgonomonas Eubacterium Actinobaculum Lactobacillus Tessaracoccus Moraxella Coprococcus Butyricimonas Demetria Johnsonella Parvimonas Butyrivibrio

78 The relative abundance of bacterial V3 tags (at Phylum level) F:Female, M:Male, C:with caries, N: no caries S:Saliva, P: plaque Ling et al., Microbial Ecology, 2010, 60(3):677-90

79 Clustering of the 8 oral bacterial communities Saliva and dental plaque habored distinct bacterial communities.

80 Heatmap of 72 predominant bacterial genera

81 6 genera are associated with dental caries significantly (p < 0.05) Ling et al., Microbial Ecology, 2010, 60(3):

82 Acknowledgement Zhiqiang Hu Kaiye Cai Peng Jia Ben Jia Prof. Liping Zhao, Shanghai Jiao Tong University Prof. Chunsheng Xiang, Zhejiang University Zongxin Lin, Zhejiang University