1. Discovery tools for human genetic variations
Gabor T. Marth
Department of Biology
Boston College
Chestnut Hill, MA 02467

2. Sequence variations
Human Genome Project produced a reference genome sequence that is 99.9% common to each human being
sequence variations make our genetic makeup unique
SNP
Single-nucleotide polymorphisms (SNPs) are most abundant, but other types of variations exist and are important

3. How do we find variations?
comparative analysis of multiple sequences from the same region of the genome (redundant sequence coverage)
diverse sequence resources can be used
EST
WGS
BAC

4. Steps of SNP discovery Sequence clustering Cluster refinement
Multiple alignment
SNP detection

5. Computational SNP mining – PolyBayes
1. Utilize the genome reference sequence as a template to organize other sequence fragments from arbitrary sources
Two innovative ideas:
2. Use sequence quality information (base quality values) to distinguish true mismatches from sequencing errors
sequencing error
true polymorphism

6. SNP discovery with PolyBayes
genome reference sequence
1. Fragment recruitment (database search)
3. Paralog identification
2. Anchored alignment
4. SNP detection

7. Sequence clustering
Clustering simplifies to search against sequence database to recruit relevant sequences
Clusters = groups of overlapping sequence fragments matching the genome reference
genome reference
fragments
cluster 1
cluster 2
cluster 3

8. (Anchored) multiple alignment
The genomic reference sequence serves as an anchor
fragments pair-wise aligned to genomic sequence
insertions are propagated – “sequence padding”
Advantages
efficient -- only involves pair-wise comparisons
accurate -- correctly aligns alternatively spliced ESTs

9. Paralog filtering The “paralog problem”
unrecognized paralogs give rise to spurious SNP predictions
SNPs in duplicated regions may be useless for genotyping
Challenge
to differentiate between sequencing errors and paralogous difference
Sequencing errors
Paralogous difference

10. Paralog filtering
Pair-wise comparison between fragment and genomic sequence
Model of expected discrepancies
Orthologous: sequencing error + polymorphisms
Paralog: sequencing error + paralogous sequence difference
Bayesian discrimination algorithm

11. Paralog filtering

12. SNP detection Goal: to discern true variation from sequencing error
polymorphism

13. Bayesian-statistical SNP detection
polymorphic permutation A C T G
monomorphic permutation
Bayesian posterior probability
Base call + Base quality
Expected polymorphism rate
Depth of coverage
Base composition

14. Priors Polymorphism rate in population -- e.g. 1 / 300 bp
Distribution of SNPs according to minor allele frequency
Distribution of SNPs according to specific variation
Sample size (alignment depth)

15. SNP score
polymorphism
specific variation

16. Validation – pooled sequencing
African
Asian
Caucasian
Hispanic
CHM 1

17. Validation -- resequencing

18. Properties of SNP detection algorithm
frequent alleles are easier to detect
high-quality alleles are easier to detect

19. The PolyBayes software
First statistically rigorous SNP discovery tool
Correctly analyzes alternative cDNA splice forms
Available for use (~70 licenses)
Marth et al., Nature Genetics, 1999

20. SNP mining: genome BAC overlaps
SNP analysis
overlap detection
inter- & intra-chromosomal duplications
known human repeats
fragmentary nature of draft data
candidate SNP predictions

21. BAC overlap mining results
~ 30,000 clones
>CloneX
ACGTTGCAACGT
GTCAATGCTGCA
>CloneY
ACGTTGCAACGT
GTCAATGCTGCA
25,901 clones
(7,122 finished, 18,779 draft
with basequality values)
21,020 clone overlaps
(124,356 fragment overlaps)
ACCTAGGAGACTGAACTTACTG
507,152 high-quality candidate SNPs
(validation rate 83-96%)
Marth et al., Nature Genetics 2001
ACCTAGGAGACCGAACTTACTG

22. SNP mining projects
1. Short deletions/insertions (DIPs) in the BAC overlaps
Weber et al., AJHG 2002
2. The SNP Consortium (TSC): polymorphism discovery in random, shotgun reads from whole-genome libraries
Sachidanandam et al., Nature 2001

23. Genotyping by sequence
SNP discovery usually deals with single-stranded (clonal) sequences
It is often necessary to determine the allele state of individuals at known polymorphic locations
Genotyping usually involves double-stranded DNA  the possibility of heterozygosity exists
there is no unique underlying nucleotide, no meaningful base quality value, hence statistical methods of SNP discovery do not apply

24. Genotyping
homozygous peak
heterozygous peak