![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
The Hidden Provision in the Big Ugly Bill that Makes Trump King
There's more linkage here.
I genuinely do not have the energy to read all of this. I will be sending out an email to my senators, I guess.
I have a reputation - not undeserved, either - as a skeptic of AI approaches in drug discovery. There is just so much hype, and so much misunderstanding. But that doesn't mean that I am reflexively hostile to computational and machine-learning approaches in general. There are too many counterexamples where it really has proven useful. But as I've said many times, these tend to be in areas with (1) a relatively well-bounded set of questions you're hoping to answer and (2) a large amount of high-quality experimental and physical data for your ML approach to work with. By "high-quality" I mean gathered under conditions which are as internally consistent as possible and which include both positive and negative data (when appropriate to the experimental design).
Materials science is an area that can fit these criteria in a very attractive way, and there accordingly have been a lot of ML efforts on metal alloys, ceramics, polymers, metal-organic structures, etc. The experimental space in these areas is gigantic, and if you can get a machine learning approach to point you to the more productive parts of it that's a big advantage indeed. Unfortunately, the hype is well established here as well, and there is no more egregious current example than this preprint from an MIT student that came out last fall. The manuscript detailed the experiences at a large unnamed R&D company doing material science research, and found that the scientists using AI assistance were significantly more productive and produced demonstrably more innovative materials that in turn led to more patent filings. A lot of commentary ensued, with plenty of it using the paper's findings to cheerlead AI in general: here was proof, real proof, that AI-driven science was the way of the future and that the future was now. We were just all going to have to get used to it. Interestingly, one of the results that got widely noted was that the already-highest-performing scientists were the ones that got the most benefit from adding AI assistance, although they reported lower job satisfaction as they realized just how much of their work it was able to do for them. You can see how this made a splash.
Those first two commentary links (from the Wall Street Journal and The Atlantic) make for interesting reading now, because the MIT student in question features prominently in both of them, speaking fluently and in detail about the work he did, the results laid out in his paper, and their implications. But it now appears that it may well have all been made up. Every last bit. MIT says that it received allegations about the paper that prompted an investigation, and that the two professors who were cited in the preprint now say that "we want to be clear that we have no confidence in the provenance, reliability or validity of the data and in the veracity of the research". Even at the time, some eyebrows had been raised about the numbers in the paper: just what company was this work done at? Why would they allow an MIT undergraduate so much leeway inside their R&D operation? How were these new materials assessed for novelty, anyway? How could you have so many people involved with only one person to coordinate all the data and author the manuscript? But many people wanted to believe, and to have their existing beliefs confirmed so throroughly.
So what's the real state of the art for AI in materials science? It's running a bit behind the preprint's futuristic take, that's for sure. Here's a paper from a group at Ottawa looking at the metal-organic framework (MOF) field, which would seem ripe for ML approaches. You have a huge range of metals to work with, an even larger list of possible scaffolding materials, and depending on the conditions you use (solvent, temperature, pH, additives) you can even get completely different crystal structures out of a single metal/scaffold pair. God knows that there are many people trying to get a handle on the broader questions in the field, but there are so many of those (synthesis, structure, stability, all sorts of possible use cases), and much of the literature still tends to be in the descriptive "Hey, here's some more of the darn things" mode. It's a huge intruiging mess, and I can tell you from personal experience that it's a lot of fun to work in. But it really does need all the help it can get.
The Ottawa group takes a close look at several open-access databases of MOF structures, which people use to try out their machine learning approaches on. And unfortunately they have found that these are in bad shape. There are way too many listed structures that have metal atoms in the wrong oxidation states (and in some cases, actually impossible ones). The paper describes "alarming error rates exceeding 40% in most databases" and you can imagine what you get when you shovel data of that quality into machine-learning algorithms. Yep, models that are state-of-the-art crap. They go on to show that using these for computing new MOFs gives you structures that themselves are wrong at least half the time, which means that the aforementioned crap is being used to produce still more scientific debris. The authors do identify some lower-error-rate databases (none of which seem to be the most popular), and the whole paper should be a wake-up call to the entire community of computational MOF researchers to slow down a bit and clean things up before hitting the ol' Start button again. That might be useful advice in general. . .
Over at For Better Science, Leonid Schneider has a fascinating look into the papers-for-cash schemes that have been going on for years now in the scientific literature. A pharmacology professor at Univ. Fribourg in Switzerland, Csaba Szabo, was contacted by one of these shady outfits with an offer, or rather, a whole selection of offers. "Ms. Kristen" from the "Alliance Academy of Science" (whatever the hell that is) wrote Prof. Szabo in2023 with an "invitation for potential cooperation", and he was able to work that interaction into his new book Unreliable: Bias, Fraud, and the Reproducibility Crisis in Biomedical Research, which was published earlier this year. Remarkably, the same people came back around to him (this time in a very similar message from "Julia" at the "Anzen Academy of Science Research"), and this time he and Schneider teamed up to see just how this would work and how far they could take it.
As you'll see from the link above, Anzen says that they operate out of China and that "we are a client service company, most of our clients are in our country, their English level is weak, but you are experienced, we would like to invite you to help our clients complete the paper writing and submission, And we can offer compensation" When Szabo asked for more details (all of these interactions were over WhatsApp chats), he was told that they had two main branches:
One is research project cooperation. The other is fast publishing cooperation.
Type 1: Researchers give topics or titles under clients need, then conducting the experiments.
Type 2: Show clients researchers’ ready titles which the experiments are ongoing or finished and clients choose them.
Fast publishing cooperation also has two types:
Type 1: Evaluating articles, evaluating the quality of articles, suggesting journals, and providing several comments.
Type 2: The language editing, scientific and structural improvements, submission and follow-up till acceptance.
As you would imagine, he needed more clarification after these statements, and what emerged was a detailed schedule of potential payouts that depended on the amount of "cooperation" and the impact factors of the journals where the resulting articles would appear, as well as the speed of acceptance at the journals themselves. It was a long list of variations on the themes of ghostwriting, authorship manipulation, and outright bribery to get things published in actual journals. These things, depressingly, have been going on for a long time now. This list was extremely similar to one he'd gotten from "Ms. Kristen" before; these are obviously the same outfit. As you'll see, they quickly starting sounding him out about journals where he had any editorial control or pull for the "fast pubishing cooperation" end of things.
And indeed, on indicating his ability to help with all these issues, Szabo was sent a manuscript with eight Chinese authors from Qingdao University on the ever-popular topic of some long noncoding RNA and its effects. Well, it's an ever-popular one in manuscripts from Chinese hospitals, as a look through the literature will immediately confirm. There's an endless list of these things, you can adduce most any downstream effects you feel like, the figures needed can (if necessary) be whipped up out of thin air pretty easily, and odds are excellent that no one will ever pay attention to your paper or ever try to reproduce anything in it. The For Better Science post shows that these authors already have extensive experience with duplicated images in their other publications; this is not their first time around the track.
Anzen also has a long list of people helping them out who have been around that track a few times, too. Szabo managed to get them to send a list of recent payouts by expressing skepticism about whether they were really delivering and whether he'd get paid for his end of the deal, and you can see that there's a lot of activity. Some of the names on the list are traceable to Iranian researchers who have suspiciously wide-ranging publication records, and it's clear that Anzen has recruited plenty of willing helpers in exchange for cash. And as he says, "It’s clear that any academically corrupt individual — particularly one with editorial connections — could easily “place” dozens of these Anzen products into indexed journals and collect a handsome side income in the process"
And this, friends, is the state of the scientific literature. I should note that while this is a serious problem, and has long been a serious problem, it is even worse now than ever. Consider the sorts of people who populate the current administration here in the US and their attitudes towards federally funded research. Do we really need to give them more cans of gasoline to throw on the fires that they're setting in their attacks the quality and usefulness of the research that they're funding? A literature that's increasingly polluted with junk is not only no good to anyone, it's a source of real harm, and it lends itself to all sorts of bad-faith attacks on scientific research in general. Those of us who do actual work and generate actual results, who write real papers and want to be able to read other real papers when we turn to the scientific literature - we need to speak up more about this situation and we need to start cleaning house more vigorously.
I wrote here in 2022 about some truly startling results from a clinical trial in patients with severe, debilitating lupus. They were treated through T-cell therapy, with modified T cells that were engineered to go after particular other cells of the immune system - in this case, the B cell plasmablasts that were the constant source of the "autoantibodies" directed against the patients' own tissues. This was accomplished by sending the T cells after CD19-positive B cells, which in effect "resets" the B-cell landscape once they have done their destructive job. The patients, who had disease so severe that it was destroying their lives, for which no other therapies were available, and whose lupus was of a type that no clinician had ever seen improve on its own. . .appear to have been cured. They are in drug-free remission, and no one until now has ever seen anything like it.
That result, which was the culmination of years of attempts by researchers around the world to accomplish such a B-cell reset, attracted a tremendous amount of attention in the immunology field, as well it should have. Now this overview at Nature reports that there are at least 85 therapies involved in about 380 clinical trials following up on this idea (!) The article bins these into three types. There are autologous cell therapies (like the above, where a patient's own cells are engineered for a personal treatment), with 95 clinical trials underway. And there are allogenic cell therapies, which are "off the shelf" cells that are not derived from the patients under treatment (43 trials) and non-cell therapies as a catch-all category (233 trials!) The cell-based modalities are using T cells (of various types) and NK (natural killer) cells, while the non-cell trials are (mostly) multispecific antibodies of different types. CD19 is a popular target, as you'd figure, but CD20 has even more trials going (antibodies against this one have shown varying amounts of efficacy with relapses being common), while others target CD38, CD22, and more possible antigens.
The diseases that are being targeted include multiple sclerosis and lupus with the most trials, with a long tail of less-common autoimmune diseases after that (ANCA-associated vasculitis, generalized myasthenia gravis, idiopathic thrombocytopenic purpura, and many more). There are readouts due later this year, and plenty more to come after that. So far (and it's early) I don't think anyone has had quite as spectacular a result as was seen in those severe lupus patients, but immunology is a very, very large field that we frankly do not understand well enough to make a lot of solid predictions in. I would expect some combinations of treatment modes, targets, and diseases to produce some truly useful results, while others that might have looked just as reasonable could show much lower response rates for reasons that aren't clear - or at least not yet.
And that's the thing about this onslaught of clinical work: we could end up with completely new treatments for some of these conditions (or for subsets of patients therein), but no matter what we're going to learn a tremendous amount about the role of B cells in their etiologies, from several different angles all at once. Which is just what's needed in a field as multifaceted as autoimmune disease. We're not going to get all the answers - God only knows when we're going to have all the answers about the immune system - but we will know a great deal more than we know now. Good luck to all involved!
As a medicinal chemist, most of the candidate compounds I’ve made have been solids. Some of them start out looking like oils and syrups, but often enough when they’re clean enough those will crystallize on standing. I have had some low-melting solids, to be sure, and those can be a bit of a challenge to formulate if you get serious about them as drug candidates. But you’d figure that solids might be pretty straightforward, right?
Unfortunately, not always right. There are solids and there are solids, on a sliding scale from “beautifully crystalline the same every time” to “amorphous stuff”. Mixed all along that scale is the constant problem of polymorphs, different crystalline forms of the same compound which can be formed under varying conditions. I’ve written about polymorphs quite a few times here; they’re a well-known pitfall because of their potential for different solubilities, bioavailability, and so on. Of course, if your initial crystal form turns out to be too high-melting and difficult to dissolve in anything, maybe you’d rather search for another form that’s more tractable! Then there are solvates, solid forms that contain solvent molecules as part of their crystalline units. The most common are hydrates (with water molecules involved), but all sorts of other solvents can do this too, although you might not feel good about ingesting many of them as drugs. Some of these crystals are very stable, while others can break down on heating or other treatment.
At the noncrystalline end, the term “amorphous” is thrown around pretty loosely by most of us. Sensu strictu it refers to a solid with no crystalline order at all, just a disordered heap of molecules. But in practice, close examination of what appear to be non crystalline amorphous drug substances often reveals a mixture of extremely small crystals, with perhaps several different polymorphs present all at once, and these can be mixed in with particles that do seem truly amorphous.
Traditionally, this situation has usually been treated as something to be avoided. The advantages of having a single crystallines form are numerous: you have a constant standard to compare to in properties like melting point, and the formation of such crystals tends to be a purifying step all in itself. Through careful optimization, you can standardize both the crystal type and constrain the particle size as they form. Amorphous solids, though, have the reputation of being difficult to reproduce exactly and could prove tricky to compare batch-to-batch. But on the good ol’ other hand, amorphous solids can dissolve much more easily than crystalline ones, and with some drugs you need all the help you can get.
So how do you get a real amorphous solid? One route is to cool/concentrate the substance in a way that it just doesn’t have time to get its crystalline affairs in order. Spray drying is a well-known technique that can give you this effect. Another is lyophilization (freeze-drying) where you start with an aqueous solution, freeze it solid, and then remove the water by subliming it off directly into the vapor state under high vacuum. Similarly, some solvate crystals can have their solvent molecules removed in such a way that a new crystalline state never gets a chance to form. The physical manipulations of milling/micronizing/drying can break a crystalline solid down into an essentially amorphous one, too, but if you’re not careful you can arrive at conditions that let crystalline order return (for example, if things heat up too much and cool down relatively slowly).
This new paper does a really good job reviewing the history of such questions and how amorphous compounds have made a comeback in recent years. This has come along with a renewed appreciation that such solid forms are essentially metastable - that is, there are almost certainly lower-energy crystalline states out there (one polymorph or another!) that are thermodynamically favored compared to the amorphous state. So it’s going to be your job to avoid any conditions that will give your amorphous solid an escape route. That will likely involve temperature and mechanical handling (as just mentioned) as well as relatively humidity (since amorphous substances almost always take up more water from the atmosphere than crystalline ones do). Similarly, they are also more exposed to atmospheric oxygen, so if you have some instabilities in that direction (which is never a good feature) your problems are going to be worse than ever in the amorphous form.
The paper has copious references to studies with existing drugs that are (or have been) produced in amorphous form, along with what instrumental methods are best suited to characterize them. There’s a particular emphasis on amorphous solid dispersions, which are mixtures of the drug substance and some polymeric carrier. These have been the subject of a lot of work in the last couple of decades, since they promise a way to deliver the advantages of the amorphous form but with greater stability on storage and handling. As we get into larger-and-stranger structural space with modern treatment modalities, these issues are surely going to grow in importance!