Here's an idea I had towards the end of my last postdoc, but only played around with it a little bit. It relates to the pharmacophore search software Pharmer.
I've mentioned David Koes's Pharmer here before. For the purposes of this discussion the key point is that it can do ultra-fast pharmacophore searching. The catch is that you need to generate conformers and index them in advance, but once done the search software whizzes through them. For example, check out the ZINC Pharmer applet over at the Camacho Lab's website.
Once something becomes ultra-fast, it opens up new possibilities for its use. In particular, we can consider using an ultra-fast pharmacophore search tool to discover pharmacophores in the first place.* So imagine tuning the pharmacophore definition by searching against a set of inactives seeded with actives; you could have a GA handling the tweaking of the pharmacophore and so on.
This method has a couple of nice features. If you are interested in selectivity (think of the 5-HT2A versus 5-HT2C I mentioned in an earlier installment) you could penalise matches against molecules known to bind to only the wrong target and reward matches to molecules known to bind to only the right one. This was why I was interested in this method, but in the end I didn't have enough data to proceed down this road; I did put Pharmer through its paces though and tried some basic optimisations of pharmacophore definitions using this approach.
In short, as far as I know no-one has ever carried out pharmacophore discovery in this way, so it could be something interesting to try especially in the context of developing a selective pharmacophore.
* I think this possibility is mentioned at the end of the Pharmer paper. But as far as I remember I thought of this independently. Makes no difference in any case.
No comments:
Post a Comment