ComPath Comparative Metabolic Pathway Analysis Tool

1
ComPath Comparative Metabolic Pathway Analysis
Tool

• Kwangmin Choi and Sun Kim
• School of Informatics
• Indiana University

2
Contents

• Introduction
• System Components
• Current Implementation
• Experiment Result
• Future Plan

3
INTRODUCTIONSYSTEM COMPONENTS
4
Introduction

• ComPath is a web-based sequence analysis system
  built upon
• KEGG (Kyoto Encyclopedia of Genes and Genomes)
• PLATCOM (A Platform for Computational Comparative
  Genomics)

5
KEGGKyoto Encyclopedia of Genes and Genomes

• Four Databases
• PATHWAY  32,657 pathways generated from 262
  reference pathways
• GENES 1,213,035 genes in 32 eukaryotes 260
  bacteria 24 archaea
• LIGAND 13,387 compounds, 2,543 drugs, 11,161
  glycans, 6,446 reactions
• BRITE 7,817 KO (KEGG Orthology) groups
• KEGG adopts EC enzyme classification system

6
EC system 0/2

• An Old, but still universally accepted system by
  biochemists
• EC system was developed long before protein
  sequence or structure information were available,
  so the system focuses on reaction, not sequence
  homology and structure
• Many biochemists and structural biologists try to
  harmonize newly available chemical, sequential,
  and structural data with traditional
  understanding of enzyme function.

7
Problems in EC system 1/2

• Inconsistency in the EC hierarchy
• For each of the six top-level EC classes,
  subclasses and sub-subclasses may have different
  meanings.
• e.g. EC1. are divided by substrate type, but
  EC5. by general isomerase type
• Problem with Multi-functional enzymes and
  multiple subunits involved in a function
• EC presumes only a 111 relationship between
  gene, protein, and reaction.
• Different sequence/structure, but similar EC
• Two enzymes with lower sequence identities
  sometimes belong to the same or very similar EC.
• e.g. o-succinylbenzoate synthase across several
  bacteria have below the 40 sequence identity

8
Problems in EC system 2/2

• Similar sequence/structure, but different EC
• Even variation in the fourth digit of the EC
  number is rare above a sequence identity
  threshold of 40.
• However, exceptions to this rule are prevalent.
• e.g. melamine deaminase and atrazine
  chlorohydrolase have 98 identical, but belong to
  different EC.
• No information on sequence/structure-mechanism
  relationship
• EC system considers only overall transformation
• Similarity among sequences is strongly correlated
  with similarities in the level of a common
  (structural domain-related) partial reaction,
  rather than overall transformation
• How to combine enzyme structure data with partial
  reaction data?
• Research Goal
• We provide a computational environment for enzyme
  analysis via genome comparison
• And it will be built on PLATCOM system

9
Our Research Goal

• We provide a computational environment for enzyme
  analysis via multiple genome comparison
• And it will be built on PLATCOM system

10
PLATCOMA Platform for Comparative Genomics

• Providing a platform for comparative genomics ON
  THE WEB
• Comparative analysis system for users to freely
  select any sets of genomes
• Scalable system interactively combining
  high-performance sequence analysis tools

11
CURRENT IMPLEMENTATION
12
ComPath

• ComPath KEGG PLATCOM
• Not just for retrieving information from
  Database,
• but focuses on analyzing enzymes using the
  enzyme-genome table
• Easy to use
• Optional Upload a user sequence and/or a saved
  enzyme-genome table data
• Select a metabolic pathway
• Select any combination of genomes in KEGG
• Create an enzyme-genome table
• Then use the table for various enzyme sequence
  analysis tasks

13
Screenshot Pathway Selection

• 11 categories
• 123 pathways
• Users can upload the previous Enzyme-Genome table
  datatype to continue analysis

14
Screenshot Genome Selection

• 250 genomes from KEGG database
• Users can select genomes by taxonomical and
  alphabetical order

15
Enzyme-Genome Table

• An enzyme-genome table allows for tests on
  whether a certain enzyme in a given pathway is
  present or missing using sequence analysis
  techniques.
• Information in this table can be easily saved,
  uploaded, transferred.
• Users also can upload their sequence set, e.g.,
   an entire set of predicted proteins in a newly
  sequenced genome, and perform annotation of the
  sequences in terms of KEGG pathways.

16
Screenshot KEGGs Ortholog Table STATIC!
17
Screenshot ComPath Enzyme-Genome Table
INTERACTIVE!
18
Screenshot Upload Query Genome and Table
Editing Functions
19
Sequence Analyses

• Missing enzyme search
• Pairwise (FASTA) and multiple sequence alignment
  (CLUSTALW),
• Domain search using SCOPEC/SUPERFAMILY and PDB
  domains
• Domain-based analysis using hidden markov models
  (HMM),
• Contextual sequence analysis (currently not
  available)
• Sequence analysis for further investigation
• Phylogenetic analysis of enzymes in selected
  genomes,
• Gibbs motif sampler.
• BAG clustering
• Contextual sequence analysis (currently not
  available)

20
Screenshot Sequence Analysis Functions
21
TEST
22
ExperimentsGenomes, Queries, Pathways

• Selected Genomes
• B.subtilus, B.Halodurans, E.coli
• H.Influenza, H.pylori, M.genitalium, Y.pestis KIM
• Query genomes
• M.tuberculosis
• A.aeolicus
• B.anthracis
• Metabolic Pathways
• 00010 (glycolysisglycogenesis),
• 00020 (TCA cycle)

23
ExperimentsComparison of Sequence Analysis
Methods

• Four methods (abbr.)
• HMMer
• HMM search using the whole sequence
• CSR
• HMM search using common shared regions generated
  by BAG program
• SCOPEC
• Domain search using SCOP/SUPERFMAILY and PDB
  database
• FASTA
• Simple FASTA search
• Cutoff
• 1e-10, 1e-20, 1e-30

24
ExperimentsOverall Design
25
Screenshot ComPath Enzyme-Genome Table
INTERACTIVE!
26
Experiment Results (e.g.)
Query Genome Pathway Mehthod Sensitivity Specificity E-value
M. tuberculosis Path 00010 HMMer 0.596491228 0.454545455 1.00E-30
CSR 0.666666667 0.454545455 1.00E-30
SCOPEC 0.614035088 0.348484848 1.00E-30
FASTA 0.649122807 0.378787879 1.00E-30
HMMer 0.623188406 0.524590164 1.00E-10
CSR 0.739130435 0.360655738 1.00E-10
SCOPEC 0.652173913 0.418032787 1.00E-10
FASTA 0.811594203 0.204918033 1.00E-10

Query Genome Pathway Method Sensitivity Specificity E-value
M. tuberculosis Path 00020 HMMer 0.535714286 0.769230769 1.00E-30
CSR 0.642857143 0.846153846 1.00E-30
SCOPEC 0.535714286 0.769230769 1.00E-30
FASTA 0.678571429 0.615384615 1.00E-30
HMMer 0.516129032 0.777777778 1.00E-10
CSR 0.709677419 0.666666667 1.00E-10
SCOPEC 0.548387097 0.777777778 1.00E-10
FASTA 0.741935484 0.333333333 1.00E-10
27
A. aeolicus
28
B. anthracis
29
M. tuberculosis
30
FUTURE PLAN
31
Future PlanMore Resources

• ComPath is being extended to incorporate more
  resources, including
• KEGG LIGAND A composite database consisting of
  compound, glycan, reaction etc.
• ProRule A new database containing functional
  and structural information on PROSITE profiles
• SFLD Structure-Function Linkage Database
• Also we are developing databases and algorithms
  for enzyme analysis, e.g. Classifiers using a
  database of enzyme-specific HMMs.
• ComPath is in an early stage of system
  development and we solicit feedback and
  suggestions from biology and bioinformatics
  communities.

32
Future PlanMore Algorithms and Tools

• More integrative understanding on biochemical
  network evolution
• Algorithms to handle isozyme problem
• Algorithms to computationally reconstruct
  alternative pathways
• Algorithms to combine sequence, structure,
  chemical reaction, and contextual information for
  better enzyme annotation
• Etc.
• ComPath is in an early stage of system
  development and we solicit feedback and
  suggestions from biology and bioinformatics
  communities.