Novel Peptide Identification using ESTs and Sequence Database Compression

1
Novel Peptide Identification using ESTs and
Sequence Database Compression

• Nathan Edwards
• Center for Bioinformatics and Computational
  Biology
• University of Maryland, College Park

2
What is missing from protein sequence databases?

• Known coding SNPs
• Novel coding mutations
• Alternative splicing isoforms
• Alternative translation start-sites
• Microexons
• Alternative translation frames

3
Why dont we see more novel peptides?

• Tandem mass spectrometry doesnt discriminate
  against novel peptides......but protein
  sequence databases do!
• Searching traditional protein sequence databases
  biases the results towards well-understood
  protein isoforms!

4
Novel Splice Isoform
5
Novel Splice Isoform
6
Novel Mutation
Ala2?Pro associated with familial amyloid
polyneuropathy
7
Novel Mutation
8
Searching ESTs

• Proposed long ago
• Yates, Eng, and McCormack Anal Chem, 95.
• Now
• Protein sequences are sufficient for protein
  identification
• Computationally expensive/infeasible
• Difficult to interpret
• Make EST searching feasible for routine searching
  to discover novel peptides.

9
Searching Expressed Sequence Tags (ESTs)

• Pros
• No introns!
• Primary splicing evidence for annotation
  pipelines
• Evidence for dbSNP
• Often derived from clinical cancer samples

• Cons
• No frame
• Large (8Gb)
• Untrusted by annotation pipelines
• Highly redundant
• Nucleotide error rate 1

10
Other Search Strategies

• Genome Corrected ESTs
• Large (2Gb)
• Controls for nucleotide error rate
• Polymorphism lost, potential errors introduced
• Genome Clustered ESTs
• Small, Gene model
• Convergence to well-understood isoforms
• Controls nucleotide error rate
• Full-Length mRNAs
• Incomplete gene coverage, most are already in
  IPI

11
Other Search Strategies

• Genome
• Large (6Gb), lots of non-coding DNA
• Find novel ORFs, no sampling bias
• Miss spliced peptide sequences.
• Genscan Exons
• Small, find novel ORFs.
• Miss spliced peptide sequences.
• How should we interpret peptide identifications
  with no mRNA evidence?

12
Compressed EST Peptide Sequence Database

• For all ESTs mapped to a UniGene gene
• Six-frame translation
• Eliminate ORFs lt 30 amino-acids
• Eliminate amino-acid 30-mers observed once
• Compress to C2 FASTA database
• Complete, Correct for amino-acid 30-mers
• Gene-centric peptide sequence database
• Size lt 3 of naïve enumeration, 20774 FASTA
  entries
• Running time 1 of naïve enumeration search
• E-values 2 of naïve enumeration search results

13
Compressed EST Peptide Sequence Database

• For all ESTs mapped to a UniGene gene
• Six-frame translation
• Eliminate ORFs lt 30 amino-acids
• Eliminate amino-acid 30-mers observed once
• Compress to C2 FASTA database
• Complete, Correct for amino-acid 30-mers
• Gene-centric peptide sequence database
• Size lt 3 of naïve enumeration, 20774 FASTA
  entries
• Running time 1 of naïve enumeration search
• E-values 2 of naïve enumeration search results

14
SBH-graph
ACDEFGI, ACDEFACG, DEFGEFGI
15
Compressed SBH-graph
ACDEFGI, ACDEFACG, DEFGEFGI
16
Sequence Databases CSBH-graphs

• Original sequences correspond to paths

ACDEFGI, ACDEFACG, DEFGEFGI
17
Sequence Databases CSBH-graphs

• All k-mers represented by an edge have the same
  count

1
2
2
1
2
18
CSBH-graphs

• Quickly determine which k-mers occur at least
  twice

2
2
1
2
19
de Bruijn Sequences

• de Bruijn sequences represent all words of length
  k from some alphabet A.
• A 0,1, k 3 s 0001110100
• A 0,1, k 4 s 0000111101011001000

20
de Bruijn Graph A 0,1, k 4
1
1
0
1
0
1
1
0
1
0
1
0
1
0
0
0
21
Correct, Complete, Compact (C3) Enumeration

• Set of paths that use each edge exactly once

ACDEFGEFGI, DEFACG
22
Correct, Complete (C2) Enumeration

• Set of paths that use each edge at least once

ACDEFGEFGI, DEFACG
23
Patching the CSBH-graph

• Use artificial edges to fix unbalanced nodes

24
Patching the CSBH-graph

• Use matching-style formulations to choose
  artificial edges
• Optimal C2/C3 enumeration in polynomial time.
• Chinese Postman Problem
• Edmonds and Johnson, 73
• l-tuple DNA sequencing
• Pevzner, 89
• Shortest (Common) Superstring
• MAX-SNP-hard, 2.5 approx algorithm

25
C3 Enumeration
in-out
in-out
Cost k
26
C3 Enumeration
in-out
in-out
Cost 0
Cost 0
Cost k
27
Reusing Edges

• ACDEHAC, ACDFHAC, ACDGHACD

28
Reusing Edges

• C3 ACDEHACDFHAC, ACDGHACD

29
Reusing Edges

• C2 ACDEHACDFHACDGHAC

30
C2 Enumeration
in-out
in-out
4
10
Shortcut paths
7
31
Implementation

• CSBH-graph construction
• Determine non-trivial nodes directly
• Consecutive non-trivial nodes determine edges
• C3/C2 enumeration
• C3 Trivial assignment of artificial edges
• C2 Depth-first search Goldbergs CS2
  min cost flow code
• Eulerian path algorithm
• Can be applied to entire EST database
• Condor grid and PBS cluster for CSBH-graph
  construction
• Large memory machine for C3/C2 enumeration

32
Conclusions

• Peptides identify more than just proteins
• Compressed peptide sequence databases makes
  routine EST searching feasible
• Currently available for download
• Can include other sources of peptide sequence at
  little additional cost.
• CSBH-graph edge counts C2/C3 enumeration
  algorithms
• Minimal FASTA representation of k-mer sets

33
Acknowledgements

• Chau-Wen Tseng, Xue Wu
• UMCP Computer Science
• Catherine Fenselau, Crystal Harvey
• UMCP Biochemistry
• Calibrant Biosystems
• PeptideAtlas, HUPO PPP, X!Tandem
• Funding National Cancer Institute