Using the Fisher kernel method to detect remote protein homologies

1
Using the Fisher kernel method to detect remote
protein homologies

• Tommi Jaakkola, Mark Diekhams, David Haussler
• ISMB 99
• Summarized by O, Jangmin (2001/09/22)

2
Abstract

• Detecting remote protein homologies
• Fisher kernel method
• Variant of Support Vector Machines using new
  kernel function
• Derived from Hidden Markov Models

3
Introduction (1)

• Detecting protein homologies (sequence-based
  algorithm)
• BLAST, Fasta, PROBE, templates, profiles,
  position-specific weight matrices, HMM
• Comparison by (Brenner 1996 Park et al. 1998)
• SCOP classification of protein structures
• Remote protein homologies existing between
  protein domain in the same structural
  superfamily.
• Statistical models like PSI-BLAST and HMMs are
  better than simple pairwise comparison methods.

miss many important remote homologies
4
Introduction (2)

• Generative statistical models (HMMs)
• Extracting features from protein sequences
• Mapping all protein sequences to points in a
  Euclidean feature space of fixed dimension.
• General discriminative statistical method to
  classify the points.
• Improvements acquired
• Over HMMs alone.

5
Methods

• How generative models work. (HMMs)
• Training examples ( sequences known to be members
  of protein family ) positive
• Tuning parameters with a priori knowledge
• Model assigns a probability to any given protein
  sequence.
• The sequence from that family yield a higher
  probability than that of outside family.
• Log-likelihood ratio as score

null model
6
Discriminative approaches

• Using both positive and negative examples
• Parameter is tuned so that the model can
  optimally discriminate members of the family from
  nonmembers.
• When training examples are few
• Likelihood ratio is optimal if generative models
  perfectly fit to data but
• Discriminative methods often performs better.

7
Kernel methods

• Discriminant function L(X)
• Where Xi, i 1,,n and hypothesis class H1,
  H2
• the sequence of the family, - outside of
  the family
• Contribution of Kernel
• ?i overall importance of the example Xi.
• Measure of pairwise similarity K(Xi, X)
• User supplies the type of kernel for the
  application area!!

8
The Fisher kernel (1)

• Deriving kernel function from generative models
• Advantage 1 handle variable length protein
  sequences!!
• Advantage 2 encoding of prior knowledge about
  protein sequences
• HMMs (difference)
• Kernel function specifies a similarity score for
  any pair of sequences.
• Likelihood score from an HMM only measures the
  closeness of the sequence to the model itself.

9
The Fisher kernel (2)

• Sufficient statistics
• Each parameter in HMM Posterior frequencies
• Of particular transition.
• Of generating one of the residues of the query
  sequence.
• Reflects the process of generating the query
  sequence from HMM.
• Alterative of sufficient statistics Fisher
  score
• Magnitude of the components how each
  contributes to generating the query sequence.

10
The Fisher kernel (3)

• Kernel function used in this paper.
• note that its fixed vector.
• Summary
• Train HMM with positive examples.
• Map each new protein sequence X into a fixed
  vector, Fisher score.
• Calculate the kernel function
• Get resulting discriminant function (SVM-Fisher)

11
The Fisher kernel (4)

• Combination of scores
• There might be more than one HMM model for the
  family or superfamily of interest.
• Average score
• Maximum score

12
Experimental Methods

• Methods
• SVM-Fisher (this paper)
• BLAST (Altshul et al. 1990 Gish States 1993)
• HMMs using SAM-T98 methodology (Park et al. 1998
  Karplus, Barrett, Hughey 1998 Hughey Krogh
  1995l 1996)
• Measurement of recognition rate for members of
  superfamilies of the SCOP protein structure
  classification (Hubbard et al. 1997)
• Withholding all members of SCOP family
• Train with the remaining members of SCOP
  superfamily
• Test with withheld data
• Question Could the method discover a new family
  of a known superfamily?

13
Overview of experiments

• Database
• SCOP version 1.37 PDB90 consisting of protein
  domains, no two of which have 90 of more residue
  identity
• PDB90 eliminates redundant sequences.
• Generative models
• SAM-T98 HMMs (alignment of the domain sequence
  and final set of homologs)
• Data selection
• Get 33 test families from 16 superfamilies.
• Evaluation strategy
• Assessing to what extent it gave better scores to
  the positive test examples than it gave to the
  negative test examples.

14
SCOP a Structural Classification of Proteins
database

• Hierachical levels
• Family clustered proteins by common evolutionary
  origin residue identities of above 30, lower
  sequence identities but very similar functions
  and structures
• Superfamily low sequence identities but probably
  common evolutionary origin
• Fold same major secondary structure in the same
  arrangement and with the same topological
  connections

15
Figure 1 Separation of the SCOP PDB90 database
into training and test sequences, shown for the G
proteins test family
16
Multiple models used

• Modeling superfamily
• SAM-T98 starts with a single sequence (the
  guide sequence for the domain) and build a model
• Using a subset of PDB90 for building model of
  other families (Too many sequences)
• Train SVM-Fisher method using each of models in
  turn

17
Details on the training and test sets

• All PDB90 sequence outside the fold of the test
  family were used as either negative training or
  negative test examples.
• Reverse test/training allocation of negative
  examples, and repeat experiments.
• Fold-by-fold basis split of negative examples.
• For positive examples
• PDB90 sequences in the superfamily of the test
  family are used.
• Homologs found by each individual SAM-T98 model
  are used.

18
BLAST methods

• WU-BLAST version 2.0a16 (Althcshul Gish 1996)
• PDB90 database was queried with each positive
  training examples, and E-values were recorded.
• BLASTSCOP-only
• BLASTSCOPSAM-T98-homologs
• Scores were combined by the maximum method

19
Generative HMM models

• SAM-T98 method
• Null model reverse sequence model
• Same data and same set of models as in the
  SVM-Fisher
• Combined with maximum methods

20
Results

• Metric the rate of false positives (RFP)
• RFP for a positive test sequence the fraction
  of negative test sequences that score as good of
  better than positive sequence.

21
G-proteins

• The result of the family of the nucleotide
  triphosphate hydrolases SCOP superfamily
• Test the ability to distinguish 8 PDB90 G
  proteins from 2439 sequences in other SCOP folds.
• Table 1
• In SVM-Fisher
• 5 of the 8 G proteins are better than all 2439
  negative test sequences.
• Maximum RFP
• Median RFP
• Figure 2
• RFP curve

22
Figure 1 Separation of the SCOP PDB90 database
into training and test sequences, shown for the G
proteins test family
23
Table 1. Rate of false positives for G proteins
family. BLAST BLASTSCOP-only, B-Hom
BLASTSCOPSAMT-98-homologs, S-T98 SAMT-98, and
SVM-F SVM-Fisher method
24
Figure 2 4 methods on the 33 test families.
Curve of median RFP
25
Discussion

• New approach
• to recognition of remote protein homologies make
  a discriminative method built on top of a
  generative model (HMMs)
• Discriminative method on top of HMM methods
• Significant improvement
• Combining multiple score would be improved.
• Allocation problem
• Different training set for tuning HMM and
  different training set for discriminative model
• Extend the method to identify multiple domains
  within large protein sequences