23 April 2011


The sequencing of the human genome has thrown up lots of potential targets, but choosing which ones to pursue is difficult: many are not biologically relevant and many are shaped such that small molecules are unable to affect their activity. “Druggability” is a popular neologism that captures both of these ideas; it refers to whether a protein can be targeted by a small molecule – preferably an orally bioavailable one – to treat a disease. However, the two components of druggability are really separate concepts, and in this month’s issue of Drug Discovery Today Fredrik Edfeldt, Rutger Folmer, and Alex Breeze coin a new term – “ligandability”. A protein is ligandable if potent small-molecule ligands can be found for it. Obviously for a protein to be druggable it needs to be ligandable, and thus it would be nice to assess this characteristic as quickly as possible. How can this be done?

Enter fragments. Because fragments have lower complexity than lead-sized (let alone drug-sized) molecules, hit rates from fragment screens tend to be higher. If a binding pocket exists in a protein, a small library of just 1000 fragments or so should produce a good range of hits. In fact, Phil Hajduk and colleagues at Abbott found several years ago that fragment screens predict the success of lead discovery campaigns. In the new paper, Edfeldt and colleagues, all at AstraZeneca, analyzed 36 internal discovery projects where both fragment screens and HTS had been conducted. They used data from the fragment screens to categorize targets into three ligandability bins:
  • Low: low hit rate, best affinities > 1 mM, low diversity of hits
  • Medium: intermediate hit rate, best affinities 0.1 – 1 mM, some diversity of hits
  • High: high hit rate, best affinities < 0.1 mM, high diversity of hits
Remarkably, all 12 targets with a low ligandability score failed HTS. Of targets that scored medium or high ligandability, 17/24 were successful in HTS screens, and 20/24 were advanced into hit-to-lead studies. These successes include targets such as BACE1 (medium ligandability), which failed HTS but which led to potent leads using fragment-based approaches. Of course, a ligandable protein may still not be druggable if it is ultimately not essential for a disease, but you often don’t discover this until after years of clinical trials.

AstraZeneca is now using fragment-based ligandability screening to help assess which targets to pursue: those with low ligandability are only pursued when the biology is truly compelling. On the flip side, targets that have failed conventional HTS but have high ligandability are reexamined using alternative hit discovery techniques, such as fragment-based methods. This seems like an appealing approach: fragments not only help drug hunters avoid throwing out the baby with the bathwater, but also to avoid drowning in dirty bathwater. I wonder how many other companies are using similar strategies.


Darren Begley said...

Hi Dan - This is a nice survey of AZ's internal FBLD, and like Phil Hajduk's referenced paper, underlines the importance (and advantages) of conducting real experiments as part of target assessment. We tried to use computational tools to predict the likelihood of what we would call "crystallographic ligandability" (see Doug's MIE book chapter). But this method was far from perfect, and never as informative as simply trying new targets out with a subset of our library.

Question: what is the "fast-follower" strategy?

- db

Dan Erlanson said...

Hi Darren,

I agree; nothing beats empiricism!

"Fast follower" is sometimes also called "patent busting." The idea is to take published patent applications from competitors, make some of the best compounds, and then med-chem into novel chemical space. This can be very effective: the current HIV protease inhibitors, for example, are far more effective than the first generation compounds.