Development of a random background to understand ligand optimization

As a working medicinal chemist, we have often started a project with new screening hits against a new target. If we are lucky, there may be some SAR in some analogues that were screened at the same time or we can simply buy further compounds to test, and see “what we get”. I am confident many of you have thought we should be doing a complete scan of every position of the hit molecules, but instead, drive forward against the straight forward synthetic chemistry. Many of us, thinking to improve med-chem understanding, have thought that the right thing to do is performing a ‘methyl scan’ of the series, but with pressure of resources and timeline have not done so.

The authors of this paper studied systematic scanning of every position of a set of ‘hit compounds’ with methyl, chloro, fluoro, aromatic C-H to N and OH (see abstract picture above), and then testing binding affinity and early in-vitro DMPK assays to explore what happens. For most projects, with a new series, there is a requirement to make a base set of compounds with a bank of data to understand what is going on. Can potency and ADMET targets be made? Hence the title in the paper ‘random background to understand ligand optimisation’. The approach the researches took deliberately made small perturbations ‘without design and without a bias towards improvement’ to yield a set of 257 compound aimed at five protein targets. This provides a full dataset that is free from reporting bias, which makes analysis of literature data limited.

The paper is well written and a very clear report of the results is made. A surprise was the finding that 11% of the compounds made had a 10-fold improvement in binding / potency. Analysis of the ADMET data has clearly shown the usually med chem problem of increasing potency at the cost of other properties, even for small perturbation such as these. The authors described this a ‘frustrated landscape’. Whilst difficult, it does provide the base set to find paths through the multi-parameter optimisation. The authors note the low improvement in metabolic stability with the substitution of a fluoro atom (See also – “A statistical analysis of in vitro human microsomal metabolic stability of small phenyl group substituents, leading to improved design sets for parallel SAR exploration of a chemical series.” Dossetter, Alexander G. Bioorganic & Medicinal Chemistry (2010), 18(12), 4405-4414.)

The authors are not the first to describe such an approach and the discussion covers many of these, including The Topliss Tree, ‘magic methyl’ effects and “may smack of the ‘methyl, ethyl, propyl, butyl, futile’ approach”. The work is quality food for thought, and many research programs should consider this approach of baselining their new series.

For some years our software MCPairs has had the “hit-to-lead” option in RuleDesign® which will generate this set of compounds and a bit more. It is based on the most common changes as found by Matched Molecule Pair Analysis. This work shows this might be useful in generating the first set of compounds to make for a new series.

To read the full article, click here