Recherche de variants génomiques en oncologie clinique. Avec des diapos, données & scripts R de: Yannick Boursin, IGR Bastien Job, IGR

Transcription

1 Recherche de variants génomiques en oncologie clinique Avec des diapos, données & scripts R de: Yannick Boursin, IGR Bastien Job, IGR

2 Génétique constitutionnelle At hospital Blood sample Sequence gene panel Look for specific alteration (BRCA) Research Genotype or Sequence Compare disease and healthy cohorts GWAS studies, 1000 Genome Project

3 NGS dans le diagnostic de génétique familiale BRCA1/2 (breast/ovary cancer) XPC, XPV.. (melanoma) ERCC1 (colorectal cancer) 10/19/16 Yannick Boursin 3

4 Génétique somatique Finding somatic mutations in the tumor genome Normal tissue (blood) Tumor tissue Extract and sequence DNA Look for mutations, structural variations Repeat for several patients Look for recurrent events Warning: ,000 SNVs between a tumor and normal tissue from the same person (genome-wide)

5 NGS in precision medicine trials Clinical trials: MOSCATO (GR), SAFIR (GR), SHIVA (Curie), Ipilimumab (anti-ctla4), Nivolumab (anti-pd1), Trastuzumab (anti-her2), Cetuximab (anti-egfr) 10/19/16 Yannick Boursin 5

6 Sequencer quoi? Panel de gènes Une série d exons d intérêt (gènes de cancer= 100kb) Exome Tous les exons du génome (30 Mb) Whole genome Le génome complet (3 Gb)

7 Un pipeline «Variants»

8 Un vrai pipeline «variants» 10 Reads (Fastq) Reference genome (Fasta) Variant Selection Manual Variant Annotation Anovar Quality Control FastQC Mapping Bowtie2 Variant calling Varscan Somatic Index Samtools Mpileup Target Regions (Bed) Target Intersection Intersect Bam Preprocess #1 GATK local realignment Preprocess #2 GATK QC recalibration

9 FastQC Metrics Look at the different metrics for both reads Problem: the per base sequence quality of the Read2 are quite lowtowards the end Solution: Trim the 25bp from the 3 end of the reads Ø Higher confidence in the sequenced information (FastqTrimmer)

10 Illumina Illumina run C Illumina run C Illumina run D Illumina run D

11 PGM PGM run A PGM run A PGM run B PGM run B

12 Trimming and discarding low quality reads A first Quality Control of raw reads is mandatory and can be established according to the application ('N', adapter sequences, barcode, contamination, etc.) ACTGATTAGTCTGAATTAGANNGATAGGAT GATCGATGCATAGCGATCAGCATCGATACG CGGCGCTCCGCTCTCGAAACTAGCATCGAC ACTGAC AGCATCAGGATCTACGATCTAGCGAACTGAC ACTGAC ACTACTTACGACATCGAGGTTAGGAGCATCA ACTAGGCATCGGCATCACGGACNNNNNNNN ACTAGCTATCGAGCTATCAGCGAGCATCTATC ACTAGCTACTATCGAGCGAGCGATCATCGAC CTGACTACTATCGAGCGAGCTACTAACTGAC ACTGAC ACTATCAGCTAGCGCTTCAGCATTACCGT ACTANNGACTAGGAATTAGCTACTGAGCTAC ACTAGCAGCTATATGAGCTACTAGCACTGAC ACTGAC NNNNNNNNNNNNNNNNNNNNNNNNNNNNN 10/19/16 Yannick Boursin Processed reads: blue parts are to be kept, green and red parts to be 12 removed

13 Reads (Fastq) Reference genome (Fasta) Variant Selection Manual Variant Annotation Anovar Quality Control FastQC Mapping Bowtie2 Variant calling Varscan Somatic Index Samtools Mpileup Target Regions (Bed) Target Intersection Intersect Bam Preprocess #1 GATK local realignment Preprocess #2 GATK QC recalibration 7-9 AVRIL 2014

14 Alignement des reads sur le génome de référence A G T C Programmes: Bowtie, BWA, Tophat, STAR, HiSat

15 Alignment key parameters - Repeats Approximately 50% of the human genome is comprised of repeats Treangen T.J. and Salzberg S.L Nature review Genetics 13, /19/16 NGS and Bioinformatics Yannick Boursin 15

16 Alignment key parameters Repeats 3 strategies -1- Report only unique alignment -2- Report best alignments and randomly assign reads across equaly good loci -3- Report all (best) alignments A B A B A B Treangen T.J. and Salzberg S.L Nature review Genetics 13, /19/16 Yannick Boursin 16

17 Alignment key parameters Using single or paired-end reads? The type of sequencing (i.e. single or paired-end reads) is often driven by the application. Exemple : Finding large indels, genomic rearrangements,... However, in most cases, the pair information can improve mapping specificity - Single-end alignment repeated sequence A C G A C T C A C G A C T C Reference Genome Sequence A C T A C G A C T C T A C G A G C A T C T A C G A G C T A C T A G C G A T C T A C G A G C T G C G A G C A A C G GC C A A C - Paired-end alignment unique sequence A C G A C T C G G C C A A C A C G A C T C G G C C A A C Reference Genome Sequence A C T A C G A C T C T A C G A G C A T C T A C G A G C T A C T A G C G A T C T A C G A G C T G C G A G C A A C G GC C A A C Yannick Boursin 17

18 Most popular aligners for variant analysis (support mismatched, gapped, paired-end alignment) BWA Li H. and Durbin R. (2009) Bowtie2 Langmead B, Salzberg S (2012)

19 Reads alignés sur le génome

20 Format SAM Contient les séquences alignées sur le génome Dans l idéal : chr ACGTGCGTTTGCGT chr GCGTGATGCGTAAG chr GTATGTTATATGTA

21 Dans la réalité: Format SAM

22 Reads (Fastq) Reference genome (Fasta) Variant Selection Manual Variant Annotation Anovar Quality Control FastQC Mapping Bowtie2 Variant calling Varscan Somatic Index Samtools Mpileup Target Regions (Bed) Target Intersection Intersect Bam Preprocess #1 GATK local realignment Preprocess #2 GATK QC recalibration 7-9 AVRIL 2014

23 Target intersection Comparer l alignement obtenu à la liste des positions visées par le protocole de capture Exon capturé Alignement non fiable Exon capturé

24 Le variant calling

25 Reads (Fastq) Reference genome (Fasta) Variant Selection Manual Variant Annotation Anovar Quality Control FastQC Mapping Bowtie2 Variant calling Varscan Somatic Index Samtools Mpileup Target Regions (Bed) Target Intersection Intersect Bam Preprocess #1 GATK local realignment Preprocess #2 GATK QC recalibration 7-9 AVRIL 2014

26 Why realign around indels? Small Insertion/deletion (Indels) in reads (especially near the ends) can trick the mappers into wrong alignments Ø Alignment scoring cheaper to introduce multiple Single Nucleotide Variants (SNVs) than an indel: induce a lot of false positive SNVs èartifactual mismatches Realignment around indels helps improve the downstream processing steps Formation NGS & Cancer - Analyses Exome

27 Local realignment identifies most parsimonious alignment along all reads at problematic loci 1. Find the best alternate consensus sequence that, together with the reference, best fits the reads in a pile 3 adjacent SNPs Realigning determines which is better consistent with the reference consistent with a 3bp insertion 2. The score for an alternate consensus is the total sum of the quality scores of mismatched bases 3. If the score of the best alternate consensus is sufficiently better than the original alignment, then we accept the proposed realignment of the reads Formation NGS & Cancer - Analyses Exome

28 Local realignment around indels Before SNVs After SNVs Deletion Deletion Formation NGS & Cancer - Analyses Exome

29 Local realignment around indels Formation NGS & Cancer - Analyses Exome

30 Local realignment around indels Formation NGS & Cancer - Analyses Exome

31 Three types of realignment targets Indels seen in original alignments (in CIGAR, indicated by I for Insertion or D for Deletion) Sites where evidences suggest a hidden indel (SNV abundance) Known sites: Ø Common polymorphisms: dbsnp, 1000Genomes Formation NGS & Cancer - Analyses Exome

32 Known sites Why are they important? Each tool uses known sites differently, but what is common to all is that they use them to help distinguish true variants from false positives, which is very important to how these tools work. If you don't provide known sites, the statistical analysis of the data will be skewed, which can dramatically affect the sensitivity and reliability of the results. 2. Recommended sets of known sites per tool Formation NGS & Cancer - Analyses Exome

33 Indel realignment steps/tools 1. Identify what regions need to be realigned Ø RealignerTargetCreator + known sites Intervals 2. Perform the actual realignment (BAM output) Ø IndelRealigner Formation NGS & Cancer - Analyses Exome

34 The quality scores issued by sequencers are biased Quality scores are critical for all downstream analysis Systematic biases are a major contributor to bad calls Example of sequence context bias in the reported qualities: before Formation NGS & Cancer - Analyses Exome after

35 Reads (Fastq) Reference genome (Fasta) Variant Selection Manual Variant Annotation Anovar Quality Control FastQC Mapping Bowtie2 Variant calling Varscan Somatic Index Samtools Mpileup Target Regions (Bed) Target Intersection Intersect Bam Preprocess #1 GATK local realignment Preprocess #2 GATK QC recalibration 7-9 AVRIL 2014

36 Reads (Fastq) Reference genome (Fasta) Variant Selection Manual Variant Annotation Anovar Quality Control FastQC Mapping Bowtie2 Variant calling Varscan Somatic Index Samtools Mpileup Target Regions (Bed) Target Intersection Intersect Bam Preprocess #1 GATK local realignment Preprocess #2 GATK QC recalibration 7-9 AVRIL 2014

37 Pileup format Describes base-pair information at each position Reference base chr A 108.$,$.,,.,,..,,.,.,..,.,,.,.,,.,..,,..,,.,.,.,.,,.,.,.,,,.,.,.,.,,,.,.,.,.,.,,.,.,.,.,,,,.,,.,.,..,,,,,.,.,,,,^F, Base qualities A 00 chr C 108 Number of reads covering the site (total depth).$t,.,,.t,tt,.,t.,.,t.tttt.ttt,t.t,ttt.tt,t,,.t TtTttt.,.,Tt.ttt.,T,.,.tT,,T,T,.tT,,t,TttTtT,T.,tt,,T,T,tttt^F.^F, Read bases:. /, = match on forward/reverse strand ??6??6 66=????? 6=7??=???<8@7=7=? 77?7?8??78?7????? 8 <8??88 9?8?0048 ACGTN / acgtn = mismatch on forward/reverse strand `-\+[0-9]+[ACGTNacgtn]+ indicates an indel Formation NGS & Cancer - Analyses Exome

38 Analyse du PileUp

39 Analyse du PileUp

40 Analyse du PileUp

41 Depth of Coverage Depth of Coverage = number of reads supporting one position ex: 1X, 5X, 100X... >1000X Reference Base SNV Sequencing Error Reference Genome Aligned Reads 2X 7X 5X brin + 2X brin - 17X 9X brin + 8X brin Calling Confidence +++ NGS - Applications and Analysis 100% = SNV Homozygote 50% 50% = SNV Heterozygote SNV and sequence context (errors) Formation NGS & Cancer - Analyses Exome

42 Variant Calling Factors to consider when calling a SNVs: Base call qualities of each supporting base (base quality) Proximity to small indels, or homopolymer run Mapping qualities of the reads supporting the SNP Sequencing depth: >=30x for constit ; >=100 for tumor SNVs position within the reads: Higher error rate at the reads ends Look at strand bias (SNVs supported by only one strand are more likely to be artifactual) Allelic frequency: Tumor cellularity will reduce the % of an heterozygous variant Higher stringency when calling indels (and sanger validation often needed) Formation NGS & Cancer - Analyses Exome

43 VarScan2 Mutation caller written in Java (no installation required) working with Pileup files of Targeted, Exome, and Whole-Genome sequencing data (DNAseq or RNAseq) Multi-platforms: Illumina, SOLiD, Life/PGM, Roche/454 Detection of different kinds of Germline SNVs/Indels (classical mode): Ø Variants in individual samples Ø Multi-sample variants shared or private in multi-sample datasets VarScan specificity is to be able to work with Tumor/Normal pairs (somatic mode): Ø Somatic and germline mutation, LOH events in tumor-normal pairs Ø Somaticcopy number alterations (CNAs) in tumor-normal exome data Formation NGS & Cancer - Analyses Exome

44 VarScan2 Most published variant callers use Bayesian statistics (a probabilistic framework) to detect variants and assess confidence in them (e.g.: GATK) VarScan uses a robust heuristic/statistic approach to call variants that meet desired thresholds for read depth, base quality, variant allele frequency, and statistical significance Stead et al. (2013) compared 3 different somatic callers : MuTect, Strelka, VarScan2 Ø VarScan2 performed best overall with sequencing depths of 100x, 250x, 500x and 1000x required to accuratelyidentify variants present at 10%, 5%, 2.5% and 1% respectively Formation NGS & Cancer - Analyses Exome

45 Somatic calling

46 Un variant tumoral: polymorphisme ou mutation?

47 Un polymorphisme «normal» (SNP) GERMLINE TISSUE TUMORAL TISSUE

48 Une mutation somatique GERMLINE TISSUE TUMORAL TISSUE

49 Varscan s Somatic P-value Variant Calling and Comparison At every position where both normal and tumor have sufficient coverage, a comparison is made. First, normal and tumor are called independently using the germline consensus calling functionality. Then, their genotypes are compared by the following algorithm: If tumor does not match normal: Calculate significance of allele frequency difference by Fisher's Exact Test If difference is significant (p-value < threshold): If normal matches reference ==> Call Somatic Else If normal is heterozygous ==> Call LOH Else normal and tumor are variant, but different ==> Call IndelFilter or Unknown If difference is not significant: Combined tumor and normal read counts for each allele. Recalculate p-value. ==> Call Germline

50 Format VCF

51 VarScan Tabulated Format Chrom Position Ref Cons Reads1 Reads2 VarFreq Strands Strands 1 2 Qual1 Qual2 Pvalue Map Qual1 Map Qual2 R1 + R1 - R2 + Rs2 - chr C Y % T chr G R % A chr G A % A chr G A % A Alt Cons : Consensus Genotype of Variant Called (IUPAC code): M -> A or C Y -> C or T D -> A or G or T W -> A or T V -> A or C or G R -> A or G K -> G or T B -> C or G or T S -> C or G H -> A or C or T Formation NGS & Cancer - Analyses Exome

52 Reads (Fastq) Reference genome (Fasta) Variant Selection Manual Variant Annotation Anovar Quality Control FastQC Mapping Bowtie2 Variant calling Varscan Somatic Index Samtools Mpileup Target Regions (Bed) Target Intersection Intersect Bam Preprocess #1 GATK local realignment Preprocess #2 GATK QC recalibration 7-9 AVRIL 2014

53 Different types of SNVs SNVs and short indels are the most frequent events: Ø Intergenic Ø Intronic Ø cis-regulatory Ø splice sites Ø frameshift or not Ø synonymous or not Ø benign or damaging etc... Example of SNV one want to pinpoint: Ø non-synonymous + highly deleterious + somatically acquired Formation NGS & Cancer - Analyses Exome

54 Resources dedicated to human genetic variations dbsnp and 1000-genomes Ø Population-scale DNA polymorphisms COSMIC Ø Catalogue Of Somatic Mutations In Cancer Non synonymous SNVs predictions Ø SIFT, Polyphen2 (damaging impact)... PhyloP, GERP++ (conservation) ANNOVAR Tools to annotate genetic variations Formation NGS & Cancer - Analyses Exome

55 Annovar «Annovar» annotates SNVs and Indels Takes Multi sample VCF (Tumor &+normal samples) Ø RefGene: Gene & Function & AminoAcid Change (HGVS format: c.a155g ; p.lys45arg) Ø 1000g2012apr_all: Minor Allele Frequency for all ethnies Ø ESP6500: Exome Sequencing Project Ø Ljb_all : predictions (SIFT, Polyphen2, LRT, MutationTaster, PhyloP, GERP++) ² Tabulated file Formation NGS & Cancer - Analyses Exome

56 Ng & Henikoff, Genome Res Utilise la conservation des domaines protéiques comme indication du caractère délétère d une substitution Classe en D(eleterious), T(olerant),.(unknown) effet réél: LacI

57 PolyPhen2 Adzhubei et al. Nature Methods Probabilistic classifier: Estimates the probability of the missense mutation being damaging based on a combination of seq+struct properties. Classe en: Benign, Possibly damaging, or probably Damaging

58 Coverage & Allelic Frequencies For CNV detection

59 Detection of copy-number variations Are there any copy-numberalteration (gain or lossof chomosomal regions, amplifications ) that could explain tumorigenesis? Yannick Boursin 59

60 La fraction allélique Vocabulaire: Germline/population: Allelic frequency, MAF (minor allele frequency). Par ex. dans données 1000Genomes. Somatic: Allelic fraction (mais souvent on utilise VAF or BAF: variant allele frequency) Où trouver l info? Colonne info#af dans VCF 10/19/16 Yannick Boursin 60

61 BAF Cellularité et Fréquence Allelique

62 Segmentation et fréquence allélique BAF Rratio R ratio=utilisé en CGH, =couverture en NGS Scott et al. Gene 2014

63 Les «signatures d Alexandrov» Yannick Boursin L. Alexandrov M Stratton, Nature

64 Signatures et origine des tumeurs 10/19/16 Yannick Boursin 64 L. Alexandrov M Stratton, Nature 201

65 Données exercice Données de cancer ovarien WES Normal + tumeur Original BAM: 10Go 2 fichiers tabulés: tumor-xxx : Tous les variants de la tumeur (somatique ou non) som-xxx : Les variants déclarés somatiques (par comparaison avec normal) et annotés Pipeline: Alignement bwa Post traitement bam picardtools sortsam GATK indelrealigner GATK variantrecalibration Variant Calling samtools mpileup varscan2 somatic Filtering variants somatiques ou LOH & Somatic P-value < 10^-3 Annotation SNPeff, SNPSIFT Script VCF>Tabulé

66 Fichier «questions» Atelier pratique