How pharmaceutical industry financial modelers think about rare diseases

This is an impressive article and an impressive couple. From the site's about page:

My name is Eric Vallabh Minikel and I’m on a lifelong quest to develop a treatment or cure for human prion diseases. I originally trained as a city planner at M.I.T. and was working as a software engineer and data analyst in the transportation sector when, in December 2011, I got some bad news. My wife and the love of my life, Sonia Vallabh, tested positive for a mutation (PRNP D178N cis-129M) that causes genetic prion disease, and that had claimed her mother’s life one year earlier. Sonia was 27 at the time. The mean age of disease onset for her mutation is around 50, and the mutation is highly penetrant, meaning she is exceptionally likely to develop the disease unless a treatment or cure is found.

Sonia and I set out on a quest to re-train ourselves as scientists. We both started taking night classes, reading papers, calling up scientists, going to conferences. We left our old careers and found jobs in research labs, and eventually enrolled at Harvard Medical School, where we are now PhD students in biological and biomedical sciences.

One thing I've learned from smart family and friends about medical care is that you can sometimes improve outcomes quite a lot if you apply brains and effort. I have relatives who managed to convert their dad's brain cancer outcomes from a couple of months in palliative care (which the doctors were recommending) to five more mostly-good years, by finding better surgeons, signing him up for clinical trials, and getting to the cutting edge of the research and applying it. For example they switched him to a ketogenic diet back when this was just starting to show results for cancer. These guys are taking this approach to the extreme! I hope they get the result they're looking for.

I would assume that somewhere on the bottom line there is an analysis of what the drug itself is doing. For instance, does the drug cure something or does it treat the symptoms of something? I imagine most of former are thrown out right off the bat. What we need is a system for testing, developing and selling medications that is less effected by the human condition. Perhaps, we should close the private sector and hand this off to the government or at least breakup the process by allowing the private sector to sell without the burden of testing and development? Perhaps, we as taxpayers, should be handling the burden of cost for testing and development?

Simple answer is probably not; even a lowering of standards you still need to know if it works - so still doing studies. You'd also then get more drugs like Lily's recently that in post-marketing turns out not to work, but ate of half a billion in sales, and a lot of patients who got not benefit.

There is no simple easy answer, it's a massively hard problem. The regulators have to balance benefit and risk, and for the most part do a hugely impressive job where they already have to add in judgement.

This is exactly correct. Also there is a good incentive structure in there - a well designed clinical trial program will move between phases, collect all data possible to support an application, and have all the relevant data to get a timely approval. The better you do on evidence collection, the sooner you are on the market, and more you have to work with

Much of the information about existing drugs was organized, but within textbooks (Merck manual, etc).

It had never been systematically structured and organized online (but likely internally within pharma). The data was (is) manually curated, included off-targets and potentially new targets, along with a suite of deep chemistry features and spectra that linked small molecules to their targets. In addition it was really the first place to organize biologic drugs, with their sequences (largely extracted manually from patents).

DrugBank was all part of a larger goal, which was to decipher the human metabolome. However, it turned out to be more successful that than (http://www.hmdb.ca is the current version of the human metabolome database, something I was also intimately involved in).

In terms of how we make money, we sell access to additional manually curated datasets (with the help of a bunch of NLP stuff for initial extraction and for QA). These datasets are structured for ML applications and integration into pharma pipelines or medical software. Additionally we sell access to an API that provides advanced queries useful for drug discovery, repurposing, and generally looking up drug information in a more uniform way. We focus on developer happiness, good documentation, and speed. Even just getting a drug product list from various jurisdictions, and keeping it up to date, is a surprisingly hard problem that the API solves.

However, keeping the data open and available for academic / student research, as well as publishing and updating drugs through the website is something we love. It's been nice to find a balance where we can get out of the cycle of grant funding but still offer something to the community and general public.