DESCRIPTION:

A Computer Science professor has designed an algorithms-based optimization tool for the biotech industry to use in identifying proteins. This method holds the promise of being more accurate, faster, and able to reduce the need to manually sequence the mass spectrometry data. Other advantages of the approach are that it does not require a large database and does not rely on binning methods which may not always be reliable.

A prototype implementation of the algorithm approach has been developed. On synthetic data generated to test the approach, it performs reliably, producing the expected protein chain. Quantifying the noise injected by the mass spectrograph is challenging, but BU scientists are confident that this algorithmic approach can successfully filter out this noise. The inventor is currently testing and tuning the approach on data obtained from Biology Department colleagues, and from on-line databases. Preliminary results of identification rates compare favorably to current industry methods.

The technology is computationally efficient; identifications are produced in fractions of a second on conventional personal computers--significantly faster than the data can be generated by laboratory equipment. Unlike most methods currently in use, this method does not rely on a large database of known spectra. This novel approach can, however, take advantage of such databases, to improve identification accuracy.

PATENT STATUS:

Intellectual property protection strategy is under review.