LOGICOIL (version 0.99.0)

EstimateProbability: Estimate oligomeric state score at each position of the input coiled-coil sequence

Description

Sub-function used in LOGICOIL to compute the oligomeric state score of input coiled-coil sequences.

Usage

EstimateProbability(id, seq, reg, pssm, LOGICOILfit, Model_Parameters)

Arguments

id

A string that represents the id name of the input sequence

seq

A character string of the amino-acid sequence to be predicted. Valid characters are all uppercase letters except ‘B’, ‘J’, ‘O’, ‘U’, ‘X’, and ‘Z’;

reg

A character string of register assignements. Valid characters are the lowercase letters ‘a’ to ‘g’. Register characters are not required to be in proper order and may start with any of the seven letters. It must always have the same length as the matching amino-acid sequence.

pssm

A profile scoring matrix generated from the LOGICOIL training data.

LOGICOILfit

The fitted model for LOGICOIL obtained using the original training data.

Model_Parameters

A list containing five elements, which are required to run the LOGICOIL model:

1. amino: a character vector containing allowed amino-acids.
2. register: a character vector containing allowed register assignments.
3. amino_interactions: a 87 x 2 matrix containing all amino-amino interactions considered in the LOGICOIL model.
4. register_interactions: a 87 x 2 matrix containing the matching register positions of the amino-amino interactions.
5. lag_matrix: a 7 x 7 matrix that gives the positional lag between register positions.

Value

Returns a matrix of dimension n x 4, where n is the length of the input coiled-coil sequence, and each column represents the LOGICOIL scores for antiparallel dimer, parallel, trimer and higher-order coiled coiled. For examples, the first row gives the predicted LOGICOIL scores of the first position of the input coiled-coil sequence. If a given sequence scores $f(x)==1$, then there is no evidence for a particular oligomeric state. If $f(x) > 1$ then there is strong evidence for a particular oligimeric state, and if $f(x) < 1$ then there is less evidence against a particular oligimeric state.

References

Thomas L. Vincent, Peter J. Green and Dek N. Woolfson. "LOGICOIL-multi-state prediction of coiled-coil oligomeric state", 29(1), pp69-76, Bioinformatics, (2013). DOI: 10.1093/bioinformatics/bts648

Examples

# load pssm data
library(nnet)
data(pssm)
data(Model_Parameters)
data(LOGICOILfit)

# priors obtained from LOGICOIL training set
prior <- c(0.6331,0.237,0.053,0.0769)
  
# compute test scores
cat("Estimating oligomeric state of coiled-coil sequences")
prob.oligo <- EstimateProbability("GCN4wt", 
  		"MKQLEDKVEELLSKNYHLENEVARLKKLV", 
			"abcdefgabcdefgabcdefgabcdefga", 
      pssm, 
      LOGICOILfit, 
      Model_Parameters)

# compute final summary probability for whole sequence
score <- apply(prob.oligo, 2, mean) / prior

# print output
print(score)