Purpose of sequence assembly

Transcription

1 Sequence Assembly

2 Purpose of sequence assembly Reconstruct long DNA/RNA sequences from short sequence reads Genome sequencing RNA sequencing for gene discovery But not for transcript quantification Variant discovery

3 Shear Genomic DNA chromosome sheared fragments

4 Sequence both ends of each fragment chromosome End sequences

5 Sequence both ends of each fragment chromosome End sequences

6 Align sequence reads to form contigs chromosome alignment

7 Paired ends allow linking of contigs into scaffolds chromosome contigs captured gaps scaffold In the sequence file, gaps are represented with Ns AGTCCCCTGGGAGATACGNNNNNNNNNNNNNNGATGATCAGCCGCATGAGCAG

8 Genome Assemblers

9 De Novo Genome Assembly Two major strategies: Overlap Layout Consensus Long reads 250 bp Pairwise comparison of reads to identify overlaps Eulerian paths/de Bruin graphs Short reads 250 bp Cataloging of subsequences (k-mers) Reconstruction of paths through the k-mers

10 Overlap Layout Consensus Fragment DNA Sequence fragments Compare all sequence reads in pairwise fashion Calculate number of overlapping bases Build a matrix

11 Overlap matrix

12 Determine Layout of Overlaps Examine best overlaps: Check their layout: GCATCGTG CATCGTGA 12. ATCGTGAT 20. AAGTGAAA 17. AGTGAAAC From: Computational Genome Analysis: An Introduction; Deonieret al.

13 Add new overlaps in a greedy fashion GACCGCAT ATGCGCAT GCATCGTG CATCGTGA ATCGTGAT GCGCATCG CGCAGCGC From: Computational Genome Analysis: An Introduction; Deonieret al.

14 Determine consensus sequence Consensus: GACCGCAT ATGCGCAT GCATCGTG CATCGTGA ATCGTGAT GCGCATCG CGCAGCGC GCGCATCGTGAT From: Computational Genome Analysis: An Introduction; Deonieret al.

15 OLC is computationally expensive 20 reads requires (20 x 20) 20 = 380 comparisons What about 10 million reads? An NP-complete problem

16 De Bruijn Graphs Break sequence reads into a set of overlapping subsequences of length k (k-mers) e.g. AGTTATCCG can be represented by the overlapping 3-mers: AGT, GTT, TTA, TAT, ATC, CCG Count how many times each k-mer occurs Place each k-mer at a node in a graph Make a path (edge) between nodes if their sequences overlap by k-1 (i.e. AGT ßà GTT) Assign the merged sequence to the edge (AGTT) Traverse each edge only once (or more if k-mer abundance implies a repeated sequence) Reconstruct genome from edge sequences

17 DeBruijn Graphs a C G A T A G T C G G Short-read sequencing b TGCAATG 3 GGCGTGC CGTGCAA ATGGCGT 5 CAATGGC Genome: ATGGCGT GGCGTGC CGTGCAA TGCAATG CAATGGC ATGGCGT ATGGCGTGCAATGGCGT Overlap Layout Consensus Vertices are k-mers Edges are pairwise alignments Vertices are (k 1)-mers Edges are k-mers De Bruijn Graphs c CAA 8 GCA 9 7 AAT TGC 10 6 ATG GTG 1 5 TGG CGT 2 4 GGC 3 GCG k-mers from vertices Genome: ATG TGG GGC GCG CGT GTG TGC GCA CAA AAT ATG ATGGCGTGCAATG k-mers from edges d CAA CA 9 AA GT AAT 10 5 CGT AT CG 4 GCG TG GTG 6 TGC 7 8 GCA ATG 1 GC TGG 2 GG 3 GGC Hamiltonian cycle Visit each vertex once (harder to solve) Eulerian cycle Visit each edge once (easier to solve) From Compeau et al., Nature Biotech, 2011

18 Eulerian cycles with sequencing errors a ATGG TGGC GGCG GCGT CGTG GTGC TGCA GCAA CAAT ATG TGG GGC GCG CGT GTG TGC GCA CAA AAT AATG b TGGA GGAG GAGT GGA GAG AGT ATGG TGGC GGCG GCGT CGTG GTGC TGCA GCAA CAAT ATG TGG GGC GCG CGT GTG TGC GCA CAA AAT AGTG c From Compeau et al., Nature Biotech, 2011

19 Eulerian cycle with repeated sequences CAA CA AA 13 GT AAT 14 4 CGT 8 9 AT CG ATG 1 3 GCG 7 TG TGG 10 GTG 5 6 TGC 2 GC GG GGC GCA Genome: ATG TGC GCG CGT GTG TGC GCG CGT GTG TGG GGC GCA CAA AAT ATG ATGCGGTGCGTGGCAATG From Compeau et al., Nature Biotech, 2011

20 It was the best de Bruijn Graph Assembly was the best of the best of times, best of times, it of times, it was times, it was the it was the worst was the worst of the worst of times, worst of times, it After graph construction, try to simplify the graph as much as possible it was the age was the age of the age of foolishness the age of wisdom, age of wisdom, it of wisdom, it was wisdom, it was the

21 de Bruijn Graph Assembly It was the best of times, it it was the worst of times, it of times, it was the the age of foolishness After graph construction, try to simplify the graph as much as possible it was the age of the age of wisdom, it was the

22 Reference-based assembly Useful when a high-quality reference genome sequence is available

23 Inchworm Assembles transcripts (dominant isoforms) Reports novel portions of alternative transcripts Chrysalis Clusters inchworm contigs into groups representing all isoforms for a given gene Builds de Bruijn graphs for each transcript Butterfly builds transcripts by using actual reads to trace paths through the graphs

24 Transcriptome assembly - Inchworm Paralogous genes: Gene A Contigs Gene B Alternative transcripts: Gene C Transcript 1 Transcript 2

25 Transcriptome assembly - Chrysalis Contigs

26 Transcriptome assembly - Butterfly

27 Assembly metrics No. of scaffolds/contigs Largest scaffold/contig N50 scaffold/contig size 50% of genome contained in scaffolds/contigs of size N50 L50 Minimum number of scaffolds/contigs with summed length 50% of genome Genome coverage (read coverage) Each base represented by an average of X reads

28 This Morning s Exercises Assemble a bacterial genome sequence Velvet Generate an interleaved dataset Choose a suitable k-mer range Run assemblies with different k-mer lengths Examine assembly metrics Discovar de novo Generate assembly Compare assembly metrics with Velvet Supplemental exercises Run assemblies using quality-trimmed input data Refine velvet k-mer range for optimal performance