I mentioned the other day that not everything in that Stuart Schreiber interview sounded sane to me, (although more of it does than I'd expected). The interviewer, Joanna Owens, asks him to expand on a statement he made about ten years ago: famously (in some circles, at any rate) Schreiber said that he wanted to - and thought that eventually he could - produce a small-molecule partner for every human gene.
A worthy goal, to be sure, but a honking big one, too. To his credit, though, Schreiber isn't making light of it:
". . .that challenge understates what we really want to do, which is to use small molecules to modulate the individual function(s) of multifunctional proteins, activating or inactivating individual functions as necessary. This is one of the differences between small molecules, for example, and the knockout of knowckdown technologies, where you inactivate everything to do with the protein of interest."
Note how things have appropriately expanded. There are a lot more proteins than there are genes (a lot more, given the surprisingly lowball figure for the total size of the human genome), and the number of protein activities is several times larger than that. He's absolutely right that this figure is the real bottom line. But here comes that Muhammed Ali side of his personality:
"Small molecules allow you to gain control rapidly, and can be delivered simply but, most importantly, we've shown that we can discover molecules that only modulate one of several functions of a single protein. . .(the scientific community has) identified 5000 out of the required 500,000 small molecules, which is similar to where the Human Genome Project was in year two of its 12-year journey. That might be a useful calibration - optimistically, we're ten years away."
Midway through that paragraph is where I start pulling back on a set of imaginary reins. Whoa up, there, Schreibster! Let's take the assumptions in order:
Small molecules allow you to gain control rapidly. . . Compared to transcription-level technology, this is largely correct. But the effects of small-molecule treatment often take a while to make themselves known, for a variety of reasons that we don't fully understand. The problem's particularly acute in larger systems - look at how long it takes for many CNS drugs to have any meaningful clinical effect. And these complex systems have other weird aspects, which make the phrase "gain control" seem a bit too confident. U-shaped dose-response relationships are common. Look at what you find in toxicology, where you see threshold effects and even hormesis, with large and small doses of the same substance showing opposite effects.
. . .and can be delivered simply. . . Well, when they can be delivered at all, I guess. But there more of them that come bouncing back at us than we'd like. In every drug research program I've been involved with, there are plenty of reasonable-looking compounds that hit the molecular target hard, but then don't perform in the cellular assay. You can come up with a lot of hand-waving rationales: perhaps the main series of compounds is riding in on some sort of active transport and these outliers can't, or they're getting actively pumped back out of the cell, or they hit some other sinkhole binding site that the others escape, and so on. Figuring out what's going on is an entire research project in itself, and rarely undertaken. Every time someone tells me that drug delivery is simple, I can feel my hair begin to frizz.
. . .we've shown that we can discover molecules that only modulate one of several functions of a single protein. . . True enough, and a very interesting accomplishment. But the generality of it is, to put the matter gently, unproven. It would not surprise me at all if there turn out to be many proteins whose functions can't be independently inhibited. The act of binding a small molecule to alter one of the functions would cause the other ones to change. And a bigger problem will be distinguishing these effects from the consequences of actually taking out that first function cleanly: how will you know when you've altered the system?
. . .which is similar to where the Human Genome Project was in year two. . . True, but that and forty dollars will get you an Aldrich Chemical can opener. The comparison isn't just optimistic - it's crazy. The problems that the genome sequencers faced were engineering problems - difficult, tricky, infuriating ones, but with solutions that were absolutely within the realm of possibility. Faster machines were made, with more computing power, and new techniques were applied to make use of them.
But as I've been saying, I'm not sure that the Maximum Inhibitor Library that Schreiber's talking about is even possible at all. Don't get me wrong - I hope that it is. We'll learn so much biochemistry that our heads will hurt. But its feasibility is very much open to question, to many questions, and we won't even begin to know the answers until we've put in a lot more work.