Investigators were able to estimate association preferences within a defined complex or a larger network using Baysian methods as opposed to binary data, according to PNAS paper.
The Stowers Institute’s Proteomics Center says that they have come up with a novel method to generate a probabilistic measure of proteins’ preference to associate with one another. They used normalized spectral counts derived from a series of affinity purifications analyzed by mass spectrometry (APMS).
Large-scale APMS studies have helped in the assembly and analysis of comprehensive protein interaction networks for lower eukaryotes such as yeast. The development of such networks for human proteins, however, has been slowed by the high cost and significant technical challenges associated with systematic studies of protein interactions, report the investigators.
The Proteomics Center team addressed this challenge by developing a method for building local and focused protein networks. With this computational approach, the probability for two proteins to associate was calculated from the bait-to-prey relationship alone. This was a step forward from other methods requiring systematic reciprocal bait-prey interactions or copurification of preys by a third bait, according to the scientists.
“Previous protein interaction networks built using protein mass spectrometry data were largely based on binary yes/no data; whether a protein is present in a sample or it is not,” explains Michael Washburn, Ph.D., director of proteomics and senior author on the paper.
“We were interested in quantitative proteomics approaches. We were able to develop a method to generate more information-rich networks, where the preference of two proteins to associate within a defined complex or within a larger network assembly can be estimated using Baysian probabilities.”
The work was published online January 24 in the Proceedings of the National Academy of Sciences.