Presentation is loading. Please wait.

Presentation is loading. Please wait.

Cancer Genome Assemblies and Variations between Normal and Tumour Human Cells Zemin Ning The Wellcome Trust Sanger Institute.

Similar presentations


Presentation on theme: "Cancer Genome Assemblies and Variations between Normal and Tumour Human Cells Zemin Ning The Wellcome Trust Sanger Institute."— Presentation transcript:

1

2 Cancer Genome Assemblies and Variations between Normal and Tumour Human Cells Zemin Ning The Wellcome Trust Sanger Institute

3 Outline of the Talk:  Project Background  Why De novo Assembly  The New Phusion Pipeline  Kmer Words Hashing  Relational Matrix  454 Reads and Assembly  Cancer Genome Assemblies from Solexa Reads  Variations between Cell Samples

4 ICGC- International Cancer Genome Consortium

5 Large-Scale Studies of Cancer Genomes  Johns Hopkins > 18,000 genes analyzed for mutations 11 breast and 11 colon tumors L.D. Wood et al, Science, Oct. 2007  Wellcome Trust Sanger Institute 518 genes analyzed for mutations 210 tumors of various types C. Greenman et al, Nature, Mar. 2007  TCGA (NIH) Multiple technologies brain (glioblastoma multiforme), lung (squamous carcinoma), and ovarian (serous cystadenocarcinoma). F.S. Collins & A.D. Barker, Sci. Am, Mar. 2007

6 Melanoma cell line COLO-829 Paul Edwards, Departments of Pathology and Oncology, University of Cambridge

7 Melanoma-Skin Cancer Disease

8 Sequencing COLO-829 on Illumina: Strategy Tumour or normal genomic DNA Fragments of defined size 0.2, 2, 3, 4 kb Sequencing 75 bp reads short insert 50 bp reads long insert Alignment using bwa, ssaha2 Somatic mutations Germline variants Sequencing performed at Illumina De novo Assembly

9 Read Coverage COLO-829 40X tumour 32X normal

10 Why De novo Assemblies  Reference is not complete There are hundreds of contigs in the current form of human genome reference and the sequence representation is only ~90%;  Reference is mosaic The DNA samples of the current reference were from 8 individuals, although there is a dominant individual, representing > 80%;  Limitations of alignment against reference Using read alignment, it can reliably call SNPs and short indels, where the indel length is dependent up the read length. But it is very hard to find structural variants, particularly long novel insertion elements;  Genomes without references  Loss of one haplotype in a diploid sample

11 De Bruijn vs Read overlap

12 New Phusion Assembler Solexa Reads Assembly Reads Group Data Process Long Insert Reads Supercontig Contigs PRono Fuzzypath Velvet Phrap 2x75 or 2x100

13 Repetitive Contig and Read Pairs Depth Depth Depth Grouped Reads by Phusion

14 Gap-Hash4x3 ATGGGCAGATGT ATGGGCAGATGT TGGCCAGTTGTT TGGCCAGTTGTT GGCGAGTCGTTC GGCGAGTCGTTC GCGTGTCCTTCG GCGTGTCCTTCG ATGGCGTGCAGTCCATGTTCGGATCA ATGGCGTGCAGTCCATGTTCGGATCA ATGGCGTGCAGT TGGCGTGCAGTC TGGCGTGCAGTC GGCGTGCAGTCC GGCGTGCAGTCC GCGTGCAGTCCA GCGTGCAGTCCA CGTGCAGTCCAT CGTGCAGTCCAT ATGGCGTGCAGTCCATGTTCGGATCA ATGGCGTGCAGTCCATGTTCGGATCA Contiguous Base Hash Base Hash K = 12 Kmer Word Hashing

15 Word use distribution for the mouse sequence data at ~7.5 fold Useful Region Poisson Curve Real Data Curve

16 Sorted List of Each k-Mer and Its Read Indices ACAGAAAAGC10h06.p1c ACAGAAAAGC12a04.q1c ACAGAAAAGC13d01.p1c ACAGAAAAGC16d01.p1c ACAGAAAAGC26g04.p1c ACAGAAAAGC33h02.q1c ACAGAAAAGC37g12.p1c ACAGAAAAGC40d06.p1c ACAGAAAAGG16a02.p1c ACAGAAAAGG20a10.p1c ACAGAAAAGG22a03.p1c ACAGAAAAGG26e12.q1c ACAGAAAAGG30e12.q1c ACAGAAAAGG47a01.p1c High bits Low bits 64 -2k 2k

17 1 2 3 4 5 6 … j … N 3 1 4 2 6 5 i N 41 0 0 0 0 R(i,j) Relation Matrix: R(i,j) – number of kmer words shared between read i and read j 41 37 0 0 0 0 37 0 22 0 0 0 22 0 0 0 0 0 0 27 0 0 0 27 0 Group 1: (1,2,3,5) Group 2: (4,6)

18 1 2 3 4 5 6 … j … 500 3 1 4 2 5 R(i,j) Relation Matrix: R(i,j) – Implementation Read index Number of shared kmer words (< 63) N......

19 Number of reads: 160.86 m; Total number of bases:35.9 Gb Reference genome size: 3.0 Gb; Sequencing platform: FLX&Titanium Read length:50-500 bp; Average read length:224 bp; Estimated read coverage: ~10X; Number of reads uniquely placed: 152.81 m; Ratio of uniquely placed reads:95.0%; Vector sources:Unknow Stats of 454 Reads – NA12878

20 Contigs: Total assembled bases:2.78 Gb Number of contigs: 526,437; Average contig length:5,280 Contig N50:11,000; Largest contig:85,538;Supercontigs: Total assembled bases:3.17 Gb Number of contigs: 54,487 Gb; Average contig length:58,263 Contig N50:1,122,317; Largest contig:8,015,559; Stats of The Assembly

21 Paired Reads Separated by “NN”

22 Error Bases Correction

23 Solexa reads : Number of reads: 557 Million; Finished genome size: 3.0 GB; Read length:2x75bp; Estimated read coverage: ~25X; Insert size: 190/50-300 bp; Number of reads clustered:458 Million Assembly features: - contig stats Total number of contigs: 1,020,346; Total bases of contigs: 2.713 Gb N50 contig size: 8,344; Largest contig:107,613 Averaged contig size: 2,659; Contig coverage over the genome: ~90 %; Mis-assembly errors:? Genome Assembly – Normal Cell

24 Solexa reads : Number of reads: 562 Million; Finished genome size: 3.0 GB; Read length:2x75bp; Estimated read coverage: ~25X; Insert size: 190/50-300 bp; Number of reads clustered:449 Million Assembly features: - contig stats Total number of contigs: 1,249,719; Total bases of contigs: 2.690 Gb N50 contig size: 6,073; Largest contig:72,123 Averaged contig size: 2,152; Contig coverage over the genome: ~90 %; Mis-assembly errors:? Genome Assembly – Tumour Cell

25 Alus : ~300bp LINEs : ~6000bp Deletions – Normal Cell

26 Alus : ~300bp LINEs : ~6000bp Deletions – Tumour Cell

27 Tumour Specific Indels Number of Deletions: 18,449 Number of Insertions: 15,899 The numbers seem to be more than what should be expected: 3000-4000 deletion/insertion; Experimental validation: ?

28 Acknowledgements:  Jim Mullikin  Yong Gu  Tony Cox  Elizabeth Murchuson  Erin Preasance  Mike Stratton


Download ppt "Cancer Genome Assemblies and Variations between Normal and Tumour Human Cells Zemin Ning The Wellcome Trust Sanger Institute."

Similar presentations


Ads by Google