1. CAP5510 – Bioinformatics Sequence Assembly
Tamer Kahveci
CISE Department
University of Florida

2. What is Sequence Assembly?
We can only sequence short fragments (100 – 500 bases).
How can we sequence long sequences (e.g., single chromosome can have hundreds of millions of bases) ?
Chop long sequence to many small fragments
Sequence all fragments
Put them together to construct the long sequence
Problem: Consider a long sequence S. Given a collection of subsequences (aka fragments or reads) of S, denoted with R = {r1, r2, …, rn}. Construct S from R

3. Sequence Assembly
Coverage: average number of reads in R containing a base in S.
Issues:
Errors in R
Repeats in S
Repeat

4. Assemblers De novo: No knowledge known about S.
Slow
Phusion (Mullikin & Ning 2003)
Arachne (Batzoglou et al. 2002)
CAP (Huang & Madan, 1992)
Mapping: A similar sequence to S is known.
Needs prior knowledge on S.
Shrimp (Rumble et al. 2009)

5. Phusion (Mullikin & Ning 2003)
Clipping: Remove low quality reads, clip ends.
Clustering: Group similar reads together.
Create a histogram of k-mers (k = 17)
Remove repetitive ones (13 or more occurrences)

6. Phusion (Mullikin & Ning 2003)
Clipping: Remove low quality reads, clip ends.
Clustering: Group similar reads together.
Create a histogram of k-mers (k = 17)
Remove repetitive ones (13 or more occurrences)
Keep a list for each k-mer showing the reads that contain it.
Find all pairs of reads sharing at least one k-mer
Keep the number of common k-mers for each such pair

7. Phusion (Mullikin & Ning 2003)
Clipping: Remove low quality reads, clip ends.
Clustering: Group similar reads together.
Assemble each cluster into a contig
Given a pair of reads, extend their matching k-mers
Join overlapping contigs
If two contigs share a read, try to put them together into a longer contig by splicing them first.

8. Euler (Pevzner et al. 2001)
Clipping: Remove low quality reads, clip ends.
Clustering: Group similar reads together.
Assemble each cluster into a contig
Create de Brujin graph
Each node is a k-mer
A directed edge indicates a dove tail overlap of k-1 positions
Find the Eulerian path on this graph (visit each edge once) – polynomial
Not the Hamiltonian path (visit each vertex once) – NP complete