Wednesday, July 23, 2008

Commercial QSAR modeling? Sorry, already patented...

QSAR has been patented in 2001 (US patent 20010049585).

Claim 1:
    A method for predicting a set of chemical, physical or biological features related to chemical substances or related to interactions of chemical substances using a system comprising a plurality of prediction means, the method comprising using at least 16 different individual prediction means, thereby providing an individual prediction of the set of features for each of the individual prediction means and predicting the set of features on the basis of combining the individual predictions, the combining being performed in such a manner that the combined prediction is more accurate on a test set than substantially any of the predictions of the individual prediction means.
They use averaging or weighted averaging of the individual predictions (claim 2). Oh, and just in case you think you are clever and you use 17, 32, etc individual predictions. Sorry, no luck either; you have to use way beyond 1M individual predictions according to the following claim ;)

Claim 2:
    A method according to claim 1, wherein the number of different predictions means is at least 20, such as at least 30, such as at least 40, 50, 75, 100, 200, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000, 2500, 3000, 4000, 5000, 6000, 7000, 8000, 9000, 10,000, 15,000, 20,000, 30,000, 40,000, 50,000, 100,000, 200,000, 500,000, 1,000,000.
What can I say about this? Please leave your opinion in the comments...


  1. US 20010049585 is a patent application, not a real patent. And with claims like these, it hardly has a chance to be granted.

  2. In short - $#%# idiotic

    More coherently, it seems like the bulk of the patent is focusing on protein structure prediction.

    Possibly a novel method to such predictions?

    But I still don't see this being awarded since ensemble models have been out for ages

  3. Patent applications sometimes ridiculous. We filed a patent for the drug-target prediction using side effects, and part of it was the fitting formula we show in the supplement. It had a few parameters, and so the application read something like: For values of A between 0 and 2, preferably 0.954.

    I guess that's they way patent lawyers write it... concerning the QSAR patent, there must be prior art for this.

  4. I wish I could be as sanguine as vfed about this patent being rejected. The paradigm at the US PTO is very much grant anything that doesn't outright claim to violate the laws of thermodynamics and let anyone that doesn't like it pay to litigate it later. It is well know that even in the face of well-known prior art, the examiners are disinclined to raise issues, since they are on a flat quota system of so many approvals a week. The only bright spot I can see is that it appears this application is part of a WIPO action initiated via EPO, so I am not sure that the actual examination will happen in Washington.

    In any case, anyone else feel like starting to accumulate disclosures as evidence of prior art? I agree with Rajarshi that it appears the overall goal is some sort of protein modeling tool, but if the patent gets issued with those claims intact, it's only an enterprising attorney away from trying to claim royalties on anyone using statistical methods on molecular descriptors, imho (IANAL). Academic and open-source developers will likely have to be the front-line on this, as the commercial developers would probably just agree to cross-license each other's patents and protect their current positions.


  5. > In any case, anyone else feel like
    > starting to accumulate disclosures
    > as evidence of prior art?

    Interesting, a quick search on WoS doesn't turn up any machine learning ensemble + QSAR work till 2002

  6. Egon..I previously blogged about the fact that chemical name to structure conversion was patented by Cambridgesoft a few years ago ( There was definitely prior art in place before their patent was granted. There were even products on the market as I recall..the granting of patents just doesn't make sense to me in many cases.