What does “drug-like space” mean to a medicinal chemist?
Introducing Dr. Paul Leeson! Paul is a medicinal chemist and drug discovery consultant with more than 35 years of industry experience under his belt… or should we say lab coat?!
Sygnature are thrilled to announce that Dr. Paul Leeson will be the Key Note Speaker at our upcoming Sygnature Discovery Synthetic Chemistry Postgraduate Symposium taking place on July 19th, 2022. Register to be a part of the audience any time up until Friday 15th July, but to be in with a shot of securing one of our speaker slots, you need to have your two-page research summary in by 5pm on Friday 10th June! You can register, and apply, here.
We had a quick catch up with Paul in advance of his talk on the 19th, read on below to hear what he had to say…
What are your thoughts on the growth of the Contract Research Organisation business model in our sector?
CRO’s in the past few years have moved from mostly executing ‘make and test’ with clients, providing the intellectual input and decision-making, to mostly fully-integrated offerings and especially collaboration and co-leadership across all drug discovery disciplines. CRO expertise has accordingly grown significantly and clients want to have CRO staff with experience of multiple projects working for them. The CRO model is now part of the fabric and is clearly here to stay! But it will be important for CROs to keep building their offerings and capabilities as we go forwards, while still delivering and retaining the advantages of flexibility and cost-effectiveness. CROs in the UK have more chemists, and recruit more than large Pharma! There is also increasing visibility of CRO science in the literature, though I think that this can improve further still. I think it’s important that the CRO community takes accountability for training and developing the next generation of drug discoverers.
In your opinion what are the ‘ingredients’ required to have a successful drug discovery programme?
Drug discovery is difficult! You obviously need multidisciplinary capabilities and know-how. Choosing the right target is critical, but above all, getting the job done is a team effort – in good teams ‘the whole is greater than the sum of the parts.’ Having clear goals, timely decision-making, balancing efficiency with open mindedness to different approaches, being hypothesis-driven, while ensuring everyone contributes maximally – in the lab, with ideas and problem solving…that’s all!!
Are there any emerging drug modalities which interest you?
If your target is extracellular and you can accept administration by injection, then pretty much any kind of modality that you can make, that is soluble enough and stable to metabolic clearance, including proteins and oligonucleotides, can be a drug. Among small molecules, mechanisms that can disconnect PK-PD linkage are very interesting, e.g. catalytic degraders (like Protacs), covalent and slowly-reversible compounds. Non- and un-competitive mechanisms have the advantage of not competing with endogenous ligands.
As a community, do you feel we are more conscious of molecular property space when making molecules in our drug discovery campaigns?
Differing points of view do exist, but I’d say there has been a real shift towards increased acceptance over the past 5-10 years. Measured logP/D right from the outset of a project is pretty much the norm now, along with DMPK in vitro screening. Greater awareness also shows up in the profiles of newly published hits and candidate drugs. The discussion now is just what the extent of that physical property space is, and whether we need new quality measures. And more emphasis on structure, e.g. seeking scaffolds and building blocks with potential for good DMPK.
What do you feel is the best way to use metrics in drug discovery?
Always ‘think quantitatively and analytically’ in projects – e. g. plotting your data, such as potency or clearance vs molecular property, is highly informative, and easy to do. Lipophilic ligand efficiency (LLE) especially has become widely recognised as important metric to optimise, and the property forecast index (PFI) has much going for it. No two projects are exactly alike of course, but awareness of the collected learnings from large datasets, for example marketed drugs and observations seen in successful optimisations and lead selection, gives pointers to problem solving, and what to anticipate and plan for to get success from your own project.
What does “drug like space” mean to you, and do we have to revise that when thinking about bRo5?
‘Drug-likeness’ and the associated term ‘chemical space’ are nebulous, indefinable concepts, these terms being incorrectly used far too often in my view. Using the Ro5, generated as a guide to permeability and solubility, as a ‘drug-like’ rule is frankly incorrect for many reasons, e.g. the properties used are changing in approved drugs over time. The all-pervasive impact of the Ro5 has unfortunately, I feel, held back investigation of less traditional space, dubbed ‘bRo5’. End of rant! With ‘bRo5’ molecules, new and valuable insights show that conformational flexibility, hydrophobic/hydrophilic collapse and exposed polarity are important factors to take account of in permeability. Importantly, it seems good bioavailability can be still be achieved with ‘bRo5’ compounds, even though in vitro permeability screens may give poor results. Running in vivo PK in screening mode is probably the way to go in these cases. Going to ‘bRo5’ risks make multiparameter optimisation tougher, so don’t do it unless you are certain your biological target really needs such molecules!
What is your feeling on the recent “AI boom”, and how do you think its impacting our industry?
Big topic! This been hugely hyped but for medicinal chemistry a sizable chunk of it is QSAR/predictive model building, enhanced by ML. This is pretty standard stuff now, and is essential practice. Generation of large numbers of ideas of what compounds to make are possible. Start-ups in the area have attracted significant funding and built impressive collaborations with large pharma. Most ‘AI-designed’ molecules so far appear to be on ‘follower’ projects – that would be feasible to do without these tools – with claims that the numbers of lead optimisation compounds needed is low, speeding up the discovery phase. The true challenge will be on tougher, novel targets. One issue is that the AI/ML models tend to be ‘black boxes’ lacking straightforward physical interpretation that chemists like to see, but it’s important to move on and change thinking if the benefits are to be realised, and things are moving fast. It is vital that AI/ML models are based on high quality data, and if so they should in principle help realise better, less biased decision-making. Across the industry there has never been a shortage of candidates, so the biggest benefit I feel will not be just going faster in discovery, but producing better quality candidates that will go further and improve success in clinical development. On a more pragmatic note, discovery computational tools I really like are: matched molecular pairs, as this powerful method relates directly to the thought processes of medicinal chemists; and virtual screening, both ligand and target based, which is now showing much promise and could soon supersede HTS.