IBM recently received a US Patent entitled “Method and apparatus for determining gene expression levels.” The patent covers a data-mining tool, which includes a data-structure populator that stores sets of data and a query builder. The sets of data are based on experiments using both synthesized probe arrays and spotted probe arrays. Glenn Held, a member of the physical sciences research staff in IBM’s Computational Biology Center, recently explained in an interview with BioArray News the technology covered by the patent.
Is this a bioinformatics apparatus or is IBM developing some sort of image analysis platform?
This is a method for analyzing bioarray data by modeling when comparing the intensities of different hybridized spots to incorporate effects such as binding energy and rate equations — physical principles that would affect the intensity — and use those to model the intensity and use them as an effective means of getting a measure of gene expression levels of different genes. In the case of Affymetrix arrays, coming up with a total expression level for a gene given the various spots that they have for the same gene.
Is this something new for IBM?
“It’s not a new instrument being developed. Part of writing the patent is the method of analysis, and in writing the patent it was also important to put in that this analysis would be run on a computer. The apparatus would be the computer to implement the algorithms that I’m talking about. It’s a general method of analysis that would be applicable to the range of existing gene expression platforms.
Is this part of the [email protected] development program?
[email protected] is one particular software package, and two of the people who were involved in that were also involved in developing this but it’s not specifically part of the [email protected] product. [email protected] assumes you already knew how much each gene was expressed and looks for patterns. This is a lower-level issue of trying to come up with a more accurate measure of how much of each of the genes is present based on the raw data.
How far along in development is the product?
It’s far enough along in [development] that we’re comfortable with the modeling. It’s not clear exactly in what form it will actually be a product. It will probably be something like [email protected], where it’s something that is downloadable.
Are you collaborating with any gene-expression firms on the development?
At this point it’s just an IBM internal effort.
Is there an estimate for a time of commercialization?
There is no particular timeframe on when it may become an actual product.