Monday, December 29, 2014

Frequent itemset mining?

What a great way to end the year in the literature!  I found a paper I absolutely love, closed the tab, went looking for it again and found something that might be even cooler.

This paper from Trung Nghia Vu et. al., appears to be a collaboration between labs at several institutions in Belgium and introduces (to me, at least) an interesting way of considering all the stuff in our MS/MS runs that we can't identify.  Basically?  Who cares!  Wait, that's not it.  But it is close.

The idea?  They go through their spectra and mine out the stuff they see all the time, whether they've identified it or not.  It becomes a frequent item.  Maybe it is some contaminant that we see all the time but no one has added it to the cRAP database.  If it is always there and it is always constant, why do we care whether we identify it or not?  Heck, lets go a step further and make sure we don't fragment it anymore.  Someone will one day figure out what it is, and yay for them, but let's get it out of the way.

Like everything I write in these boxes, this is a massive oversimplification.  I recommend this paper to everybody.  It is short, open access, and goes after things in a way I've certainly never thought of.  Now to find that other really cool paper....

1 comment: