His alternative, MSAlign+ utilizes “dynamic programming” and spectral alignment to identify proteins (5).
“In every algorithms class students learn that when there is [an] exponential explosion of variants to consider, you should not try to tabulate these variants but rather come up with an algorithm that makes the problem tractable.” Several algorithmic techniques can be used, he explains, including dynamic programming, “the magic trick behind many genomics (and now proteomics) algorithms.”
Edwards, who has used MSAlign+, says it is in some ways more powerful than ProSight PC, but also “a bit more research-grade.” Still, says Pevzner, MSAlign+ is fast and relatively comprehensive. “It actually discovers practically everything that ProSight PC finds, and more.” (Kelleher says that the recently released ProSight PC 3.0, addresses some of these criticisms.)
Such tools could be key to tackling the proteome. Bottom-up experts have identified large swaths of the proteome, but each protein can exist in multiple forms. Indeed, Paša-Tolić has detected some 3,000 unique histone isoforms in one cell type—representing just 30 genes.
And therein lies the strength of top-down, says Kelleher: It's not in the sheer number of identifications a study can produce, but the quality of the resulting data, the unambiguous identification and characterization of discrete protein isoforms.
Biologists, he says, are catching on.
“The more that you understand it, you start to see why top-down has and will continue to develop a following. And it's durable. It's not easy, but it's getting easier all the time.”