Download presentation
Presentation is loading. Please wait.
Published byRosamond Austin Modified over 9 years ago
1
Gena Tang Pushkar Pande Tianjun Ye Xing Liu Racchit Thapliyal Robert Arthur Kevin Lee
2
Biology background Algorithms De novo - Overlap-Layout-Consensus - De Bruijn graphs Reference Tools and Techniques Work flow & strategy Group management
3
Biology background Algorithms De novo - Overlap-Layout-Consensus - De Bruijn graphs Reference Tools and Techniques Work flow & strategy Group management
4
Flow Order TACGTACG 1-mer 2-mer 3-mer 4-mer KEY (TCAG) Measures the presence or absence of each nucleotide at any given position
5
Margulies et al., 2005 http://www.youtube.com /watch?v=kYAGFrbGl6E
6
~300nt per read 40X coverage (but widely varying, 12-80)
7
~ 2 Mb genome (1.8-2.0Mb) Mostly coding sequences ◦ Good for assembly Reference genomes of 4 closely related species ◦ H. influenzae ◦ H. parasuis ◦ H. ducreyi ◦ H. somnus
8
High degree of genomic plasticity ◦ 10% of genes in clinically isolated strains are novel In situ horizontal gene transfer Supragenome – distributed genome hypothesis Reference mapping relatively ineffective Making life difficult!!!
9
De novoReference mapping Number of large contigs 50500+ Unmapped reads130068000+ N50 contig size285003000 Bases in large contigs (Mb) 21.3
10
Assemble all of the H. hemolyticus genomes together Should give a more complete mapping because of higher coverage ◦ 40X * 5 genomes 200X coverage But … we get 3700 contigs ◦ (average of 50 for single strain assembly)
11
These data hint at rampant recombination Reference mapping relatively worthless
12
Whole-genome alignment (intra-species) ◦ On average, 27 insertions, 147 deletions (>90bp) ◦ Average length of non-matching seq = 321kb (18%)
13
Biology background Algorithms De novo - Overlap-Layout-Consensus - De Bruijn graphs Reference Tools and Techniques Work flow & strategy Group management
14
Read: a 500-900 long word that comes out of sequencer Mate pair: a pair of reads from two ends of the same insert fragment Contig: a contiguous sequence formed by several overlapping reads with no gaps. Supercontig (scaffold) an ordered and oriented set of contigs, usually by mate pairs. Consensus sequence: sequence derived from the multiple alignment of reads in a contig
15
Goal: Find the shortest common sequence of a set of reads. Input: reads {s1, s2, s3, …} Output: find the shortest string T such that every s_i is a substring of T. Comment: This is NP-hard problem, we need to use some approximation algorithm.
16
Process: (1) Calculate pairwise alignments of all fragments. (2) Choose two fragments with the largest overlap. (3) Merge chosen fragments. (4) Repeat step 2 and 3 until only one fragment is left.
17
Best one Take pairwise alignment Merge the best one Input reads
18
Comment: Greedy algorithm was the first successful assembly algorithm implemented. Used for organisms such as bacteria, single- celled eukaryotes. It has some efficiency limitation
19
This approach is based on graph theory. Assemblers based on this approach: Arachne, Celera, Newbler etc.
20
Sort all k-mers in reads (k~24) (1) Find pairs of reads sharing a k-mer (2) Extend to full alignment-throw away if not > 95% similar TACATAGATTACACAGATTACT GA | | | | | | | | | | | TAGTTAGATTACACAGATTACTAG A
21
One caveat: repeats A k-mer that appears N times, initiates N^2 comparisons. Solution: Discard all k-mers that appear more than c*Coverage, (c~10)
22
A graph is constructed: (1) Nodes are reads (2) Edges represent overlapping reads CGTAGTGGCAT ATTCACGTAG Overlap graph
23
A graph is constructed: (1) Nodes are reads (2) Edges represent overlapping reads CGTAGTGGCAT ATTCACGTAG Overlap graph
24
Terminology in graph theory: (1) Simple path--- a path in the graph contains each node at most once. (2) Longest simple path---a simple path that cannot be extended. (3) Hamiltonian path– a path in the graph contains each node exactly once. CGTAGTGGCAT ATTCACGTAG
25
Recall: Now we got several contigs(i.e. several longest simple paths) Find the multiple alignments of these contigs, and get one consensus sequence as our final contig.
26
Summary: (1) Based on graph theory (2) Eularian path: a path in a graph which visits every edge exactly once. (3) Example: Euler, Velvet, Allpath, Abyss, SOAPdenovo… (4) Eularian path is more efficient, however, in partice both are equally fast.
27
Break reads into overlapping k-mers. Example: 10bp read: ATTCGACTCC for k=5-mers: ATTCG TTCGA TCGAC CGACT GACTC ACTCC
28
Nodes: k-mers Edges: if (k-1) suffix of a node equals (k-1) prefix of a node, add a directional edge between them. ATTCG TTCGATCGAC
29
Whenever a node A has only one outgoing arc that points to another node B that has only one ingoing arc, the two nodes are merged. ATTGC TGCAT TGCAG TTGCA ATTGCA
31
In Velvet: (1) Error removal (2) Removing tips Tip: a chain of nodes that is disconnected on one end.
32
Consider two paths redundant if they start and end at the same nodes (forming a “bubble”) and contain similar sequences. Such bubbles can be created by errors or biological variants, such as SNPs or cloning artifacts prior to sequencing. Erroneous bubbles are removed by an algorithm called “Tour Bus”.
34
Algorithm for directed graphs: (1) Start with an empty stack and an empty circuit (Eulerian path). - If all vertices have same out-degrees as in-degrees - choose any of them. - If all but 2 vertices have same out-degree as in-degree, and one of those 2 vertices has out-degree with one greater than its in-degree, and the other has in-degree with one greater than its out-degree - then choose the vertex that has its out-degree with one greater than its in-degree. - Otherwise no Euler circuit or path exists. (2) If current vertex has no out-going edges (i.e. neighbors) - add it to circuit, remove the last vertex from the stack and set it as the current one. Otherwise (in case it has out-going edges, i.e. neighbors) - add the vertex to the stack, take any of its neighbors, remove the edge between that vertex and selected neighbor, and set that neighbor as the current vertex. (3) Repeat step 2 until the current vertex has no more out-going edges (neighbors) and the stack is empty.
36
C_k=C*(L-k+1)/L N50 size: 50% of genome is in contigs larger than N50 Example: 1Mbp genome Contigs: 300, 100, 50, 45, 30, 20, 15, 15, 10,… N50=30kbp (300+100+50+45+30=525>=500kbp) Note: N50 is meaningful for comparison only when genome size is the same
37
Map k-mer on the reference sequence, get a “location map”. Map each read onto the “location map” according to the k-mer. AATTGGGTTA location map of 5-ker CCCAATTGAAA AATGGTTACCA
39
Biology background Algorithms De novo - Overlap-Layout-Consensus - De Bruijn graphs Reference Tools and Techniques Work flow & strategy Group management
40
Standard flowgram format (SFF) A binary file format used to encode results of pyrosequencing from the 454 Life Sciences platform for high-throughput sequencing. a header section + read data sections
41
A summary of general information regarding the file content
42
Reads' universal accession numbers (h), sequence information (s), quality scores of basecalls (q), clipping positions (c), flowgram values (f) flowgram indices (i) the nucleotide bases + the quality scores
43
6 genomes, 6.sff files Number of reads ranges from 72548 to 391117 Reads Contigs/ Scaffold Assembler
46
GS De Novo Assembler: a software package designed specifically for assembling sequence data generated by pyrosequencing platforms De novo assembly Overlap-Layout-Consensus methodology Better deal with reads greater than 250bp in length GS Reference Mapper
47
Algorithms for de novo assembly Short read assembly (25~50bp) Using de Bruijn graphs. Applying Velvet to very short reads and paired-ends information only can produce contigs of significant length, up to 50-kb N50 length in simulations of prokaryotic data and 3-kb N50 on simulated mammalian BACs.
48
Open-source whole genome assembly software - Assemblers: Minimus2 - Validation and Visualization: Hawkeye - Scaffolding: Bambus - Trimming, Overlapping, & Error Correction
49
Celera MIRA Edena Finishing is a big challenge !
50
Sequencing errors: base pair misread, poly A… It is possible that some portions of genomes are unsequenced Identical and nearly identical sequences (repeats) can increase the time and space complexity of algorithms exponentially Gaps & errors
51
Finishing is taking contigs and yielding a complete sequence. Scaffolder orders contigs into scaffolds based on clone-mate pair information. Some assemblers have a simple quality-control method Check and manually assemble unresolved repeat regions Resequencing
52
Biology background Algorithms De novo - Overlap-Layout-Consensus - De Bruijn graphs Reference Tools and Techniques Work flow & strategy Group management
53
Plan will evolve ◦ Different beast than expected Write scripts to automate the pipeline Velvet did not work MIRA3 requires much time Edena is not optimized for 454 data
54
454 sequencing reads.sff files Newbler Newbler? Reference genome What can we use to reconcile assemblies? Merged assembly contigs? Scaffolding? How do we visualize and evaluate assemblies? Convert to.fasta Make all reads equal sized Velvet
55
454 sequencing reads.sff files Newbler Newbler? Reference genome What can we use to reconcile assemblies? Merged assembly contigs? Scaffolding? How do we visualize and evaluate assemblies? Convert to.fasta Make all reads equal sized Velvet
56
Human and chimp = 99% sequence similarity H. influenzae and H. influenzae = 80% s.s. H. influenzae and H. hemolyticus = ??? (<80%)
57
454 sequencing reads.sff files Newbler MIRA3 Add a third? AMOScmp Reference genome Minimus2 Merged assembly contigs Scaffolding: Bambus Visualize and evaluate: Mauve, Hawkeye, amosvalidate Specialized algorithm
59
MIRA3 Hash tags Contigs for each genome Hash tag indicates repeat? BLAST ends against contigs of conspecifics BLAST ends against completed genomes Ends from different contigs match same contig in 1 or more genomes Link together No mapping or multimapping Further processing? Ends from different contigs map to similar RefSeq regions Link together No mapping or multimapping Further processing? Yes No
60
Plans change, and knowledge changes An automated pipeline is invaluable ◦ What if 30 15 contigs? Gene finding group ◦ Just re-run the scripts
61
Core genes vs unique genes Gene clustering Codon usage
62
Biology background Algorithms De novo - Overlap-Layout-Consensus - De Bruijn graphs Reference Tools and Techniques Work flow & strategy Group management
63
Some task can be divided, some are not. Complexity of communication The optimal group size is 3-6 Mini groups
64
Group meetings Wiki main page discussion page Everyone is special, thus valuable Start earlier, get closer to perfection
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.