Genome 373: Genome Assembly. Doug Fowler

Transcription

1 Genome 373: Genome Assembly Doug Fowler

2 What are some of the things we ve seen we can do with HTS data?

3 We ve seen that HTS can enable a wide variety of analyses ranging from ID ing variants to genome- wide biology! Li Ding et al. Hum. Mol. Genet. 2010;19:R188-R196

4 But We Always Need One Thing! After we get the HTS reads, there is a common first step for all these analyses. What is it? We ve just assumed that we were given one critical piece of data. What is it?

5 But We Always Need One Thing! After we get the HTS reads, there is a common first step for all these analyses. What is it? Read mapping! We ve just assumed that we were given one critical piece of data. What is it?

6 But We Always Need One Thing! After we get the HTS reads, there is a common first step for all these analyses. What is it? Read mapping! We ve just assumed that we were given one critical piece of data. What is it? The reference genome! reference genome

7 Outline De novo genome assembly introducbon State- of- the- art assembly with short reads: the De Bruijn graph Complete course evaluabons

8 Acquiring Data How would you guys go about acquiring sequencing data for genome assembly?

9 Shotgun Sequencing Genomic DNA First, isolate genomic DNA Commins, J., ToK, C., Fares, M. A. Biol. Procedures Online (2009).

10 Shotgun Sequencing Genomic DNA Generate defined- length fragments Fragment by sonication, nuclease or transposase Commins, J., ToK, C., Fares, M. A. Biol. Procedures Online (2009).

11 Shotgun Sequencing Genomic DNA Generate defined- length fragments Sequence and assemble Commins, J., ToK, C., Fares, M. A. Biol. Procedures Online (2009).

12 Shotgun Sequencing Genomic DNA Generate defined- length fragments Sequence fragments Assemble fragment sequences Commins, J., ToK, C., Fares, M. A. Biol. Procedures Online (2009).

13 Reads are the Basic Unit of Assembly All we start with at the beginning of the assembly process is a read. read

14 Reads are the Basic Unit of Assembly All we start with at the beginning of the assembly process is a read. read Read length is a key parameter is de novo assembly

15 Reads are the Basic Unit of Assembly All we start with at the beginning of the assembly process is a read. read Technology Read Length (nt) Sanger ~1,000 HTS ~100

16 Assembling a Genome Once we have reads from randomly sheared DNA, what is our next step?

17 Assembling a Genome Align the reads to find ones that have overlaps

18 Assembling a Genome Why is this a hard problem?

19 Assembling a Genome Because it s an all-by-all comparison of the reads. We ve seen how using hashing and seeds can help.

20 Assembling a Genome What s next, now that we know which pairs of reads contain overlaps?

21 Assembling a Genome contig Assemble the overlapping reads into contigs

22 Dealing with Gaps Between Contigs Contig #1 Contig #2 Gap! Now we have a problem: gaps between contigs. How can we deal with these? Hint: we have to change our experimental design

23 Paired End Reads to Connect Contigs Contig #1 Contig #2 Gap! Sequencing paired ends enables us to bridge gaps

24 Paired End Reads to Connect Contigs Contig #1 Contig #2 Gap! If we know the length between read pairs, even better!

25 Paired End Reads to Connect Contigs Contig #1 Contig #2 Gap! How would we actually do this?

26 How Would We Actually Do This? 1) Fragment genome 2) Isolate defined sizes on a gel 3) Clone into a vector (or append HTS adapters)

27 Contigs are Assembled into Scaffolds Scaffolds are large units of assembly.

28 Contigs are Assembled into Scaffolds Of course, even this strategy won t be a complete one what regions are we likely to miss?

29 Repeat Regions are Problematic Of course, even this strategy won t be a complete one what regions are we likely to miss?

30 Assessing Assembly Quality How should we assess the quality of our assembly?

31 Assessing Assembly Quality N50 is a simple statistic for assessing assembly quality

32 Assessing Assembly Quality N50 is defined as the length of the shortest scaffold at 50% coverage of the genome

33 Assessing Assembly Quality We arrange the scaffolds from biggest to smallest

34 Assessing Assembly Quality Then identify the length of the smallest scaffold needed to cover 50% of the genome (here N 50 = 30)

35 Outline De novo genome assembly introducbon State- of- the- art assembly with short reads: the De Bruijn graph Complete course evaluabons

36 Using Graphs to Represent Assembly Imagine we acquire short reads from a small circular genome

37 Using Graphs to Represent Assembly We can represent the traditional assembly process we just talked about as a directed graph where each edge represents the best alignment between two reads

38 Using Graphs to Represent Assembly Walking the graph corresponds to assembling the genome

39 Breaking Reads into k-mers ATGGCGT In practice, we break reads into short k-mers to ensure that all k-mers in a genome are represented

40 Breaking Reads into k-mers ATGGCGT ATG TGG GGC GCG CGT k=3 mers for the first read in our example

41 Breaking Reads into k-mers ATG Given a k-mer we define its suffix as the string formed by all nucleotides except the first

42 Breaking Reads into k-mers ATG Given a k-mer we define its prefix as the string formed by all nucleotides except the last

43 A Graph With k-mers as Nodes ATG TGG We connect one k-mer to another using a directed edge when the suffix of the first k-mer equals the prefix of the second k-mer

44 Using Graphs to Represent Assembly The assembled genome can be found by visiting each node once and only once

45 Using Graphs to Represent Assembly This is equivalent to the align all reads to each other and find the optimal assembly problem

46 Using Graphs to Represent Assembly Also known as finding a Hamiltonian cycle, it s computationally very difficult

47 Euler and the 7 Bridges of Koningsberg Euler showed that we can find a path that goes through all edges of a graph exactly once, provided that every vertex is equal in in/out-degree

48 Euler and the 7 Bridges of Koningsberg It also turns out that finding a Eulerian path through all edges, if it exists, is much less computationally difficult than finding a Hamiltonian path

49 Euler and the 7 Bridges of Koningsberg How can we take advantage of Euler s observation?

50 Representing k-mers as Edges ATG Recast the graph so that edges represent k-mers

51 Representing k-mers as Edges AT ATG TG A prefix and suffix are joined by an edge when they represent an observed k-mer

52 Representing k-mers as Edges AT ATG TG Finding an Eulerian path through such a graph gives us the genome assembly

53 Finding an Eulerian Path: Hierholzer s Algorithm AAT AT ATG AA TG CAA TGG CA GTG TGC GG GT GCA GC GGC CGT CG GCG Here is a graph representing the same reads we ve been working with we want to find an Eulerian path through the graph

54 Finding an Eulerian Path: Hierholzer s Algorithm AAT AT ATG AA TG CAA TGG CA GTG TGC GG GT GCA GC GGC CGT CG GCG 1) Start with any node

55 Finding an Eulerian Path: Hierholzer s Algorithm AAT AT ATG AA TG CAA TGG CA GTG TGC GG GT GCA GC GGC CGT CG GCG 1) Start with any node 2) Walk an arbitrary path of edges back to the start node

56 Finding an Eulerian Path: Hierholzer s Algorithm AAT AT ATG AA TG CAA TGG CA GTG TGC GG GT GCA GC GGC CGT CG GCG 1) Start with any node 2) Walk an arbitrary path of edges back to the start node 3) If any node has edges not part of the current path, start another walk from that node, following unused edges and returning to the node. Append this second path to the first

57 Finding an Eulerian Path: Hierholzer s Algorithm AAT AT ATG AA TG CAA TGG CA GTG TGC GG GT GCA GC GGC CGT CG GCG The algorithm is guaranteed to give an Eulerian path if one exists

58 Picking the Best Eulerian Path 10 AT 1 10 AT 1 AA TG AA TG 9 CA GG 9 CA GG GT 5 8 CG 4 GC 3 GT 4 8 CG 3 GC 7 You may have noticed that we could make multiple distinct Eulerian paths for this graph, each of which would correspond to a distinct genome assembly

59 Picking the Best Eulerian Path AA 10 AT 1 10 AT 1 TG AA TG 9 CA GG 9 CA GG GT 5 8 CG 4 GC 3 GT 4 8 CG 3 GC 7 This problem arises because, when we converted reads to k-mers we lost linkage information across reads

60 Picking the Best Eulerian Path AA 10 AT 1 10 AT 1 TG AA TG 9 CA GG 9 CA GG GT 5 8 CG 4 GC 3 GT 4 8 CG 3 GC 7 One of our reads, ATGGCGT, spans the ambiguous region

61 Picking the Best Eulerian Path AA 10 AT 1 10 AT 1 TG AA TG 9 CA GG 9 CA GG GT 5 8 CG 4 GC 3 GT 4 8 CG 3 GC 7 One of our reads, ATGGCGT, spans the ambiguous region And we can use it to pick the right path

62 The Eulerian Path Through the Graph Gives the Sequence 10 AT 1 AA TG 9 2 CA 6 7 GG GT 8 GC 3 5 CG 4 So, now we have the right Eulerian path through the graph

63 The Eulerian Path Through the Graph Gives the Sequence AA 10 AT 1 TG A 9 2 CA 6 7 GG GT 8 GC 3 5 CG 4

64 The Eulerian Path Through the Graph Gives the Sequence AA 10 AT 1 TG AT 9 2 CA 6 7 GG GT 8 GC 3 5 CG 4

65 The Eulerian Path Through the Graph Gives the Sequence AA 10 AT 1 TG ATGGCGTGCA 9 2 CA 6 7 GG GT 8 GC 3 5 CG 4

66 References De novo assembly Compeau, Pevzner and Tesler, How to apply de Bruijn graphs to genome assembly. Nature Biotechnology, 2011 Baker, De novo genome assembly: what every biologist should know. Nature Methods, 2012 Jones and Pevzner An introduction to bioinformatics algorithms Chapter 8

67 Outline De novo genome assembly introducbon State- of- the- art assembly with short reads: the De Bruijn graph Complete course evaluabons h`ps://uw.iasystem.org/survey/142151