Variagenics of Cambridge, Mass., has developed a computational method to predict which SNPs are most likely to affect the function or stability of a target protein.
The company said it is using the computational approach to speed identification of significant genetic variations in patient populations.
The method applies only to non-synonymous SNPs (nsSNPs), which are known to result in changes in protein sequence. “It gives us a way to prioritize non-synonymous SNPs in terms of their likely effects on protein function,” said Daniel Chasman, group leader of functional annotation at Variagenics.
The approach used features derived from sequence information and structural models to anticipate which amino acid polymorphisms are likely to affect protein function and stability.
“All these features are integrated into a probability that an amino acid polymorphism that derives from a SNP will have an effect on function,” Chasman said.
Once Variagenics prioritizes the nsSNPs with the computational method, it uses the information to determine which ones warrant further experimentation.
“As we enter the post-genomic era, the ability to search through the millions of SNPs now available in databases worldwide to find those few SNPs directly linked to important biological effects becomes critically important,” said Colin Dykes, vice president, research, at Variagenics. He added that the method would help the company select “minimal, informative subsets of SNPs for application in cost-effective pharmacogenomics programs with our pharmaceutical and biotech partners.”
Chasman has estimated that between a quarter and a third of nsSNPs in the human genome affect protein function. However, the ratio of nsSNPs to the total number of SNPs in the genome has not yet been determined.
The computational method is “not perfect, but it’s fairly accurate,” Chasman said. He noted that it only suggests perturbation of function based on the principles of protein structure. It doesn’t indicate whether the function is increased or decreased.
The automated method can run a typical analysis in a matter of minutes compared with manual methods that can take from hours to days. Chasman said that the method also calibrates its findings by reference to known data sets of mutations and other proteins.
“Instead of just having an expert scientist say, ‘Yeah, that one looks good,’ I now have some formalism or theoretical basis for guessing which ones are going to have an effect on function,” said Chasman.
Chasman and Mark Adams, senior director of bioinformatics at Variagenics, published a paper on the approach in the March 23, 2001 Journal of Molecular Biology.
Variagenics has no plans at this time to commercialize the method in a software implementation.