Skip to Content

Chemical News

So What Do You Get From DNA-Encoded Libraries?

Oliv Eidam and Alex Satz at Roche have published a useful paper that compares a number of the company’s DNA-encoded libraries. There are plenty of papers on these things, but comparisons and general principles are much harder to get ahold of. In this case, they’re showing the results of 16 different libraries (made via different chemistries and building blocks) against two targets, a kinase and a phosphodieserase. The smallest library had about one million compounds in it, most were in the tens of millions, and the largest had (potentially) 100 billion compounds or so (this one was a four-step synthesis scheme).

But one of the lessons that comes out of this work is that library size had no real correlation with hit rate (that whopper library, for example, had no hits at all against these two targets). The others gave compounds that, on resynthesis, were anywhere from 2 nM to double-digit micromolar hits. Of the 57 compounds that were resynthesized (with just a methyl truncating whatever linker they had to the DNA), 35 showed activity at some level. Looking at the structures, several things are clear: for one, none of these structures were to be found in the regular Roche screening deck, CHEMBL, or the patent literature. They were all novel. Their molecular weights varied between 250 and 600, but potency did not correlate at all with increasing MW or with increasing logP (most of the compounds came in between 3 and 4 in cLogP). That’s as opposed to general screening deck, where logP does indeed (and unfortunately) correlate.

And another interesting thing was that a disproportionate number of hits were actually truncates – compounds that did not go through the whole three- or four-step plan for the library, but were intermediates along the way. That would suggest that the more elaborate DNA-encoded schemes, generating larger and more complex chemical matter, may turn out to be a waste of time (or at least waste more of it than they’re worth). It may be too early to draw that conclusion, but it’s something to keep an eye on. . .

9 comments on “So What Do You Get From DNA-Encoded Libraries?”

  1. Moreno says:

    Congratulations on the nice work! These are important considerations for DNA-encoded library planning and construction!

  2. anonymous says:

    Speaking of screening….announced today. All HTS work at Merck moving from PA to Kenilworth. CVD out to California, Infectious Diseases to West Point, PA. coupled with layoffs. As the rearrangement of deck chairs continues…….

    1. ROGI says:

      Where was this announcement made? Several of my former colleagues at HTS are questioning this info,

      1. anonymous says:

        9:00 AM in N. Wales, which is being closed. Various meetings in Kenilworth starting this morning.

  3. tangent says:

    I’m curious, is it typical that of the 57 candidates, 35 (~61%) would hit on resynthesis? What are the other 22 here, hitting due to their DNA tag?

    This type of work fascinates my inner mathematician, in relation to the way that biological ligands for one target will accidentally cross-hit to another, seemingly strangely often. (My favorite is that chemicals from anaerobic soil bacteria turn out to be spectacular inhibitors of animal neurotransmission, which I doubt they have any interest in inhibiting.)

    1. Zander says:

      Hi Tangent,

      I share your fascination concerning the fact that a lot of drugs are actually quite ‘dirty’ in that they most often also bind completely different proteins with similiar affinity. I think that this phenomenon is significantly underestimated at the moment…

      PS: you have any interesting literature about this perhaps?

    2. Former Cube says:

      A 61% hit rate on resynthesis is actually pretty good by my experience. If you do the Bayesian analysis of any large screening effort you’ll see that even a false positive rate below 5% will generate a lot of false hits when run against tens to hundreds of thousands of compounds. While it’s within the realm of possibility that the DNA tag could result in authentic binding, there are also examples of compounds in libraries that were not what they were intended to be or that they degraded in storage into something unexpected. That being said, Occam’s razor suggests that the other hits were most likely screening errors.

      All of that is said in the background of not having read this specific paper closely.

      1. tangent says:

        Oh [slaps forehead] yeah, “contents not 100% as stated on label” would be one very plausible way to end up with a false positive that doesn’t go away on retest of the material. Thanks for explaining.

  4. Robert Burns W says:

    Truncates are an inescapable fact of reality when dealing with large libraries produced by multi-step routes.
    Go back to the Affymax papers and you will see examples of hits which were unintended stalled intermediates. Ah, the good old days when combinatorial chemistry was the buzzword and big pharma fools parted with their money faster than the red sea

Comments are closed.