Ramblings: April 2023
1 May 2023
Below are collected Twitter ramblings for April 2023, with additional commentary interspersed. This is also the first time I’m having to directly deal with the consequences of the Twitter-Substack feud. As of this writing, it’s still not possible to embed live tweet links into a Substack post. As a workaround, I’ve taken screen shots of the Twitter posts I want to highlight this month and then hyperlinked them back to the original tweets. Not perfect, and much more time consuming — but it works.
This month’s photo of the month was taken on the summit of Cobble Mountain at Macedonia Brook State Park in Kent, CT. Looking west, the outline of the Catskill Mountains can be seen on the horizon. Spring comes on slowly in New England. Although the temperatures were moderate, leaf-out of the trees is in very late April and early May.
Reading the date is key to understanding this one. I’d be happy to knock off to Sedona with those margaritas though.
Despite years of articles (and indeed, entire websites) describing the characteristics of high-quality chemical probes and countless entreaties to stop publishing studies that draw sweeping biological conclusions with poor-quality probes, the curcumins of the world remain. To be fair, higher tier journals have cracked down a bit, but legions of predatory journals all too willing to publish crap have stepped into the void. Those who are not pharmacology savvy will often parrot (cite) such studies as conclusive. And thus the literature freight train of faulty conclusions lurches ever forward, kept on the rails by those who don’t know or don’t care (or both).
To me, there are a couple of red flags to watch for:
Has direct target engagement been demonstrated, both in biochemical and cellular assays, and possibly also using structural biology techniques? Or is it just inferred by referring to older literature, or resting on models not backed with experimental data?
Has selectivity for the desired target been established? A kinase inhibitor without a KinomeSCAN, a degrader without proteomics, an antibody without the whole MW range of the gel shown — all of these things are sus.
Have appropriate control experiments been done? The minimum is a negative control compound with a minimal structural change that abrogates binding to the target and also abolishes expected on-target mechanistic and functional effects. Better still is another compound from a distinct structural class that hits the same target and shows the same downstream pharmacology. If not: sus.
Does the in vitro pharmacology data make sense?
Swing too high. Are the concentrations used non-insane? Specific target binders that require 10 uM concentrations to achieve pharmacological effects: sus. An x nM inhibitor in a biochemical assay that requires 100x (or 1000x!) that concentration to show a mechanistic or functional effect in cells: sus, especially if 100x-1000x is in the high micromolar range and the shift is not expected based on e.g. a Cheng-Prusoff situation. The higher the concentration used, the higher the likelihood that you’re picking up polypharmacology. Paracelsus: the dose makes the poison.
Swing too low. If a biochemical assay pegs a target IC50 at x nM and then a cellular mechanistic or functional readout reports an IC50 of x/10 nM — that’s sus. Why would a cellular proliferation assay read out potent at a concentration where the biochemical data suggests the fractional occupancy of the target is 10% or less? And that’s ignoring cell permeability, competition with native substrates leading to Cheng-Prusoff shifts, etc. — all of which will further right shift cellular potency, not left shift it. The simplest answer is: that mf is hitting other things, and your target hypothesis is bullshit. If you believe you have an exception, the burden of proof is on you to establish why.
Does the in vivo pharmacology data make sense? Did the authors bother to establish the pharmacokinetics of the compound? Was it done in the same species, sex, strain, dose, formulation, etc. that eventually gets used in the PD or efficacy study? Did the authors measure plasma protein binding and follow the free drug hypothesis? And if they did all of that, do the free drug exposures suggest sufficient target occupancy for a long enough time to drive the observed PD and efficacy? The absence of a clear answer to any of these questions: once more, sus.
Usually unknowable, but: who the heck peer reviewed this article? Imo if a chemical biology (broadly using that term) paper reports chemical structures, that paper should be peer reviewed by a card carrying medicinal chemist.
The response to these Friday career posts has been great, and highlights that there’s still a substantial gap in job hunting knowledge for folks coming out of their academic programs and onto the industrial job market for the first time. Gaps are particularly big for people who are first-gen in any sense of that word, as I was when coming through grad school. A lot of this knowledge was hard-won in my own case, so hopefully these threads provide a useful service. I remain open, as always, to coming out to any university — virtually or in-person as budgets allow — to talk about industrial careers. It’s one of my favorite things to do, and I don’t get to do it anywhere near as much as I’d like to.
Of course here, I missed probably the most obvious hiring practice: it’s all about the Benjamins. A lot of crap is tolerated or ignored as long as a PI can bring in funding. So the university can scrape off all that lovely overhead, nom nom. Perhaps the lack of this incentive in industry, where the company supplies the funding, contributes to different hiring practices. If I’d given this poll to industrial folks, “we screen out jerks” would have been an overwhelming majority answer. Now I’ll have to actually do the poll to confirm my suspicions!
I know. Cold call sales techniques are still a thing. And conference organizers are never going to stop selling their attendee lists to vendors. Truth though: I have a lot of things to do at my job, and dealing with vendors is one thing I budget zero time for. We have whole groups of people at my company who handle business development, and if you’re talking lab supplies, then it’d be better to talk with someone who, you know, still works in a lab. So when those emails come in, I block with impunity. It’s not personal, as the saying goes: strictly business.
When SciFinder comes up empty with commercial sources and published synthetic routes, this is one of those times when a medicinal chemist needs to ask: do I need that exact compound to answer the SAR question I want to interrogate, or can I do it with a similar-looking but more synthetically accessible molecule? Always strive to make the simplest molecule that will answer the question. It’s a rare day when we’ll embark on a total synthesis project because we just have to have that one compound which there’s no literature precedent for making. Despite what our process chemistry colleagues would have you believe, we don’t usually go out of our way to litter our molecules with fluorines and stereocenters just to make their lives harder.
Can’t overstate these simple points enough. I’d wager that less than 5% of new medicinal chemists joining industry from academia just pick everything up as if they’ve been doing it their whole lives. You can only learn so much from the classroom and books. Medicinal chemistry is highly experiential, and you simply need time to build up your personal database of lived experiences to lean on in the future. And to also learn to lean on your village and not try to do everything yourself.
So great to be back on the ground at my favorite conference in Orlando this year. It’s a time to catch up on cutting-edge cancer research and connect with friends old and new from across the years. As I’ve mentioned before, I rarely go to ACS meetings anymore. Not that I have anything big against ACS or the Division of Medicinal Chemistry, it’s just that I need biology and clinical inputs into my brain more than I need chemistry ones these days. And AACR indeed has a vibrant Chemistry in Cancer Research working group that’s worth checking out if you’re a medicinal chemist working on oncology targets.
Still working on this one because it’s a long piece and my attention has been pulled in a dozen different directions this year. I hope to have it out this summer though.
My “conference rules” that I published at AACR in April 2022 are evergreen. A lot of this is down to time management skills, which are often not taught in academia but become increasingly vital as you move up the ranks. The way I see it, I have a very limited window to eat lunch, and everyone else is eating lunch at the same time. Ergo, don’t spend the lunch hour waiting in line to get lunch. Buy lunch early in the day somewhere cheaper than the convention center, and then spend the time actually eating lunch.
This is the heart of any conference: dinner and social hour!
This may or may not be a certain reception, which we may or may not have been invited to.
The hour of talks and posters I missed, which can easily be recouped at my leisure thanks to online streaming/recording, pales in comparison to my physical and mental well-being. This is the biggest change I’ve made in my approach to conferences over the years, and I don’t regret it one bit. You have to take care of yourself, or all that knowledge you’re supposed to be absorbing in the lecture hall will just leak right out of your brain.
If you’re ever fortunate enough to work on a team that discovers a marketed drug, you’ll find that you’re bonded to them for life. These peeps will always be my peeps.
I try hard to avoid getting involved with debating the whole Elon-Twitter situation. I’m going about my business, interacting with the folks I want to interact with, and doing my best to leave my corner of the Twitterverse a little better than I found it. I don’t have time or energy to get into arguments about the owner’s questionable stances.
Of all the Elon companies out there though, I admit to a soft spot for SpaceX. As an American Gen X kid, I grew up in the Space Shuttle era and had the Challenger disaster imprinted into my 9 year old brain. The retirement of the Shuttle was a loss of national pride, and I never dreamed it would be so long before we started flying our own astronauts from our own soil again. And I definitely never imagined that a private company, and not NASA, would be the ones to make it happen.
There’s still a lot of science waiting to be done in our solar system. It excites me to think about what we might accomplish without weight (which in turn leads to billion dollar custom rovers made out of titanium parts) being the primary design constraint. What can we do when launch costs are 100x lower than today, and what are we doing to get ready for that future now? For starters, as an organic and medicinal chemist, I’d love to know what’s lurking in the liquid oceans of Europa and Enceladus. It seems more likely than not that SpaceX and other private companies are going to be the ones to take our payloads there in the future.
The closer you look, the wronger it gets.
This is a hot take that I expected to get me in some hot water in some corners of the med chem universe. But I’ll stand by my assertion that PK and PD are the truly hard problems in medicinal chemistry. Every successful project reaches a point where you’ll gladly trade off some of that potency for PK. Or maybe not, because all things being equal, I’ll take the series that starts with good PK over the more potent one any day of the week. I wish we devoted half as much obsession to understanding PK/PD as we do to the molecular interactions that drive potency. The discipline would be a lot better off.
I had fun making this animation using Procreate, and have plans for a few more of these down the road.
As a discipline, I think we’re nearly out of fresh things to say about BRD4 degraders. And that sets aside that it’s possibly the most easily degraded target ever discovered, so as a validation tool for your new favorite E3 ligase ligand, it’s a poor choice.
Always something satisfying about having a paper copy in hand. We nominally have the cover art too, which was designed by someone on the Nature Reviews team. It’s accurate if a little abstract.
Part of the reason, as I stated earlier in this post, that <5% of medicinal chemists hop into their first industry job and just grasp everything is that very few of us are trained in academia as medicinal chemists. The usual skill set we hire is synthetic organic chemistry. Being able to make molecules is foundational to medicinal chemistry, but it’s the beginning of the story, not the end. We’ll layer everything else you need on top on the job. We’re counting on the skills you have solving chemistry problems of one chemistry type transferring to a different type.
Some good and thought-provoking replies in the comments on this one. I’ve never felt that AI and ML tools weren’t going to impact drug discovery processes significantly. They will play a role in shaping the field going forward, probably in a big way. But saying that we’re ready for humans to take their hands off the wheel and let AI design our drug candidates and successfully get them through clinical trials is simply a bridge too far. When we cross the boundary from feasible to fantasy, from hope to hype, that’s where I get off the ride.
Thank you SciFinder for reminding me of my own work.
The people who are concerned about ChatGPT scooping up huge swaths of stuff that used to take them hours are probably right to be legitimately worried about AI outright replacing their jobs. But as anyone who’s interrogated ChatGPT about anything related to chemistry can tell you, it’s comically bad at doing any actual chemistry. So I’ll go out on a limb and say we’re safe for a while yet.
Half my feed these days.
Dirty Harry: A man’s got to know his limitations.