1. How to Build a Horse
Megan Smedinghoff

2. Background
In February 2007, Broad Institute released a draft genome of the horse (Equus caballus)
The project cost $15 million and was funded by the National Human Genome Research Institute and the National Institute of Health
300,000 Bacterial Artificial Chromosomes were provided by the University of Veterinary Medicine in Hanover, Germany and the Helmholtz Centre for Infection Research in Braunschweig, Germany

3. Horse Genome Statistics
The horse genome contains approximately 2.7 billion base pairs
The assembly was done using 6.8-fold coverage
The sequenced horse was a thoroughbred mare named Twilight from Cornell University
Twilight posing for a picture at Cornell

4. Why Sequence the Horse?
Allows scientists to study diseases that primarily affect horses such as Glanders
SNP information can be used to connect DNA to physical characteristics and explain differences between breeds
Lots of general information about mammals can be gained by looking at the horse since very few large mammals have been sequenced

5. How the Horse Genome Affects Us
There are over 80 known genetic conditions in the horse that are analogous to human disorders
Horses have some conditions traditionally found in humans such as allergies and arthritis
Having the complete horse genome helps infer the order of evolution
Horse Racing?

6. Project Proposal
Reassemble the horse genome using the Celera Assembler
Use existing UMD software to compare my assembly with the Broad assembly and produce a reconciled horse genome
Deposit the improved assembly in GenBank
Advisor: Jim Yorke

7. Introduction to Genome Sequencing
DNA target sample
SHEAR
SIZE SELECT
e.g., 10Kbp
± 8% std.dev.
Primer
End Reads (Mates)
SEQUENCE
750bp
Vector
LIGATE & CLONE
Slide courtesy of Art Delcher

8. How Genomes are Assembled
Closure
Trim the Reads
Calculate Overlaps
Build Unitigs
Build Contigs
Build Scaffolds

9. Assembly: Calculating Overlaps5’ 3’
Read A 5’ 3’
Read A 5’ 3’
Read B 3’ 5’
Read B 3’ 5’
Read A 3’ 5’
Read A 5’ 3’
Read B 3’ 5’
Read B
Compare every possible combination of reads to find every overlap of a certain length (~40bp)
Must compare forward and reverse orientation of each pair of reads
Comparisons must take into account the possibility of sequencing errors and use alignment algorithms such as Smith-Waterman

10. Assembly: Creating Unitigs
Reads
A unitig is a set of reads that have been linked together based on overlaps
A unitig has no ambiguities

11. Assembly: Creating Unitigs (cont.)
Best Buddy Algorithm for Unitig Assembly:
If the longest overlap with read A is read B and the longest
overlap with read B is read A, then reads A and B are best buddies A B C A B C D D
Read A and Read B
are best buddies
Read A and Read B
are NOT best buddies

12. Assembly: Creating Contigs
Unitig A
Unitig B
Read 1
Read 2
Read 1 and Read 2 are mates
A contig is a set of overlapping unitigs
Contigs are assembled by using mate pair information
Since we know the distance between mates and the orientation of the mates, we can infer the placement of the unitigs

13. Assembly: Building Scaffolds
Contig A
Contig B
Reads
Scaffolds are built from contigs
The orientation and approximate distances between contigs are inferred from mate pair information
When possible, the gaps between contigs are filled in with leftover sequence

14. Arachne Assembler 24-mer indexing
Any two reads that share at least one 24-mer are paired
Each pair is scored
Contigs are created by merging paired pairs
Repeat regions are avoided during contig assembly but used during scaffold assembly
Subreads are placed after scaffold assembly
Serafim Batzoglou
Arachne Author

15. Celera Assembler
Find overlaps of at least 40bp with less than 6% error
Overlaps are found using 22-mers
After overlaps are calculated, Celera does error correction using a voting algorithm
Contigs are assembled using best buddy algorithm
Scaffolds are assembled from mate pair information
Scaffold gaps are filled when possible
Gene Meyers
Former vice president
of Celera Genomics

16. Project Expectations Fall 2007 Produce Celera Assembly Spring 2008
Produce Reconciled Assembly
General Goals
Tackle the unexpected problems that accompany genome assembly
Document my work
Validate my work wherever possible

17. Validation Genome assemblies are not perfect
I plan to validate my assembly by comparing it to the current draft
I expect about 1.5% difference between the Celera Assembly and the Broad Assembly
I will use Mummer to measure similarity between genomes

18. Graphs courtesy of Adam Phillippy
Mummer
Mummer is a piece of software created by CBCB that is used to compare genomes
Mummer locates strings of at least 18bp that are present in each genome
Plotting the results makes it easy to see insertions, deletions, inversions, etc.
Graphs courtesy of Adam Phillippy

19. Implementation Details
I plan to use the Genome cluster at University of Maryland to produce my assembly
Much of my project will utilize existing software
I intend to use Perl to write any
additional scripts that may be
needed

20. Time Permitting
The University of Maryland has recently produced a lot of software for the genome assembly pipeline, much of which has not been tested on large genomes
I hope to use programs like the UMD overlapper and Figaro to see how these programs affect my assembly
Mihai Pop
James White

21. Acknowledgements
James Yorke, Aleksey Zimin, and the Genome Group for advising me on the nature of this project
Steven Salzberg, Art Delcher, and Adam Phillippy for giving lectures and producing slides on genome assembly topics
Gene Myers paper on Drosophila
Serafim Batzoglou paper on Arachne
Wikipedia