One thesis in Eisen's thread, along with a recent NY Times piece, is that the Human Genome Project (HGP) was grossly oversold with absurd hype about how it would rapidly revolutionize medicine. There's a lot of truth in the article, though it also has a number of problems -- as I pointed out, the number of innovative drugs coming from target-based approaches was greatly and artificially depressed by simply leaving out protein therapeutics and focusing on small molecules. Now, what I just said may not make much sense to many people, so I'll try to explain.
A common, though not entirely accurate split, in drug discovery approaches is phenotypic screening vs. target-based screening. In phenotypic screening, project scientists pick some biological effect and then design an assay to measure that effect. This is what a compound library is screened against. On the other hand, target-based approaches start with a specific gene product and develop a drug using that information. The HGP provided all the possible targets, so why haven't target-based approaches delivered a wealth of new drugs?
Now one problem is that it isn't always easy to identify targets from the genome. Sometimes, particularly in this era of whole genome sequencing of rare disorders, nature shows its hand by placing a mutation smack in a gene. Unfortunately, other approaches for finding targets, such as identifying genes over or underexpressed in a disease state, have not been as productive as once hoped. It is sobering to think how few of the targets of current therapeutics could be found from genomics data (targeting activated protein kinases in cancer being a major exception). Suppose you find a mutation causing a disease, what is your next step?
The first question to ask is whether the mutation activates or inactivates the gene. Activating mutations include those that modify the protein to make it more active as well as promoter mutations which increase the expression of the gene in a disease state. These are not exclusive; the EML4-ALK fusion in lung (and sometimes other) cancers pulls the double trick, generating both a constitutively active kinase (because EML4 supplies dimerization, a key mechanism for activating tyrosine kinases) as well as expression in lung. Conversely, a mutation might inactivate the protein, by directly altering the encoded protein's activity or stability or localization, or by blocking correct transcription, splicing or translation, or a multitude of other things that must go right for a protein to be correctly expressed.
This is already a big step. Since most therapeutic agents decrease the activity of proteins, it is much easier to target activated genes than inactivated ones. Conversely, and unfortunately, an awful lot of genetic disorders are due to inactivating mutations, probably because it is much easier to destroy a needed activity than to create a problematic one. So many of the biological discoveries from genomics are inherently difficult to translate directly
A key exception is if the mutated gene encodes an enzyme. In a subset of such cases (such as Gaucher's Disease or Adenine Deaminase deficiency), enzyme replacement therapy is a successful route. Unfortunately, in many cases (such as Tay-Sachs), enzyme replacement is not practical.
Okay, you don't have an inactivated enzyme. The next question is whether your activated protein target is either extracellular or secreted (and not behind the blood-brain barrier). If so, then an antibody therapeutic might be an option. Monoclonal antibodies and related tactics have been hugely successful in a number of conditions, such as Rituxan for many leukemias and lymphomas or Herceptin for breast cancer. That's one reason I find the NY Times article severely irritating (with words like "dirty pool" and "underhanded" coming into my head); it excludes protein therapeutics from the tally, and since protein therapeutics are essentially always a targeted approach, that makes target-based discovery look worse. On the other hand, an important caveat is that while these drugs came from targeted approaches, those targets predate the HGP. Indeed, due to the long timelines of clinical development, it's not easy to think of an approved therapeutic that is directly tied to the HGP, though I suspect someone with a better knowledge of the space could rattle some off.
The other major line of attack is a small molecule The catch is the vast majority of small molecules target a very small number number of protein classes, such as ion channels, nuclear hormone receptors, G-protein coupled receptors, protein kinases, proteases and a few other categories of enzymes. Many other categories of enzymes have stuck their tongue out at drugging attempts, and attempts to drug protein-protein interactions or transcription factors have mostly been a bottomless pit for drug development funds. This was perhaps a bit of honest optimism pre-HGP that has been quashed post-HGP: most of the proteome is undruggable by the major modes of therapeutics.
Now, I haven't covered above a large number of targeted therapeutic modalities, but those are (unfortunately) a set of minor players, at least to date. In particular, nucleic acid therapeutics hold out the hope, in principal, of being able to drug any target, perhaps both to downregulate bad actors or repair good ones. For example, antisense oligos seemed like a transformative therapeutic strategy when I was an undergraduate, but few such therapies have advanced far. A major problem with nucleic acid therapeutics is successfully delivering them in vivo, both from a standpoint of stability as well as getting them to desired tissues (the liver seems to soak them up; hence a focus of some companies in that space on liver diseases). RNAi has the same challenges, though perhaps the hints of success at Alnylam will spur new activity in that space. CRISPR is the new exciting player, but far to new to evaluate. Similarly, Moderna's promise of delivering mRNAs to cells is promising, but far too early-stage to judge.
Now, another approach is to not drug the gene you find, but something else in the pathway. If you understand the biology of the system, then one can try to hit another target in the system. For example, the hedgehog pathway is inappropriately activated in some rare cancers by inactivating mutations in a negative regulator. Since that isn't a target for drug intervention, hedgehog programs have targeted downstream, activating genes. If you don't know the actors in the pathway (or perhaps if you do), then a phenotypic screen could be order. However, it is also important to note that understanding targets can assist or even enable a phenotypic screen. For example, if a mutation causes a disease by destabilizing a protein, then a phenotypic screen might attempt to find proteins which reduce that destabilization.
There are a lot of strongly held views in the target-based vs. phenotypic screening debate, and I won't claim a side -- that's not my space and I don't feel qualified. But I do feel that the debate is hindered by poor information. It's also important, in my view, to recognize that the HGP has transformed large swathes of human biology by providing a common framework for organizing vast amounts of biomedical knowledge. It is hard to argue the HGP was oversold, and in particular the speed at which therapeutics and diagnostics would appear, but conversely attempts to paint the HGP as failing to impact drug discovery and medicine are equally suspect. The HGP, and especially subsequent developments such as rapid and inexpensive whole genome resequencing, are continuing to change medicine. Translating those biological insights into therapeutic advances will remain difficult, for the reasons I outline above as well as others, but I do believe academia and the biopharma industry will continue to make important therapeutic advances that were critically dependent on the determination of the human genome sequence.