Great idea. I used to always just count down so we could start Netflix at the same point, and then sync up after a bit by pausing and restarting occasionally, this is a very elegant solution!
very useful article! two other options I used before are just installing base R on EC2, or using this RStudio AMI I found from a Google search: http://www.louisaslett.com/RStudio_AMI/
We are definitely not experts in regards to pharma (neither Alice or I have experience working in pharma) but our advisors include pharma consultants and senior scientists. We're constantly engaging them and meeting with pharma execs and scientists almost daily to learn what has and hasn't worked, and have received a lot of feedback about how to bring potential new drugs to patients. There's a line between being naive and being jaded and we are trying to stick to being as realistic as possible, and value any new information or feedback, if you know anything that's contrary to our experience.
Our video definitely was geared toward a lay audience, but we tried not to use buzzwords and bullshit and to explain the science in a more accessible way. I'm curious to know what what bothered you about the video and we will try to communicate better - at least for me, conveying scientific ideas in an understandable way has been pretty challenging and I had hoped we did a good job in the video.
We definitely didn't invent systems biology for drug discovery. But Google didn't invent search engines either, I think that we have a fresh and novel approach which is really differentiated by the details.
Another thing I think is different about the industries is that while there are tons of food delivery and photo-sharing startups, there's no really effective/disease modifying treatments for common diseases such as Alzheimer's disease, frontotemporal dementia, ALS, etc, so I think there's plenty of space for any number of new approaches to be tested in the field. I feel it's more worthwhile than pumping billions more into failed antibodies against beta amyloid, as seems to be the trend nowadays.
I am a big believer in the prion-like spread hypothesis which seems to be a common thread for neurodegenerative disease, and led a study (in press) that supports this idea in Alzheimer's disease and frontotemporal dementia. Once there is a critical mass, I think it would be challenging to slow the progression of the disease, which may speak to the importance of early treatment. However, given the proven genetic risk factors for these diseases (Mendelian forms, GWAS hits) - which are related to things like UPR, microglia activation, etc - I think that there is a possibility to slow the progression of disease. We also feel the same way you do about the repositioning angle, but it is a nice way to try to get something safe to patients that can potentially be proven quickly and cheaply in a small Phase 2 trial.
In terms of the expression data for diseases, we are doing a systemic review and re-analysis of publically available raw micoarray data from GEO, as well as RNA-sequencing and microarray data that we will publish soon and make publically available. That part is what I mentioned in the video; rather than just looking for differentially expressed genes (which is what most of the associated studies have done), we use network methods we developed to identify what we think are the key drivers of disease. Also, by combining data we get more power to find signal. To match these with drugs, we use a combination of proprietary gene expression data as well as data from large public projects.
The 1000X number should only be attached to "cheaper", and comes from in vivo and in vitro validation experiments we have done; our drug hit rate was almost 1000X higher than high-throughput screens from the literature, and our total spend to get to our one drug lead was almost that many times cheaper than what we learned that pharma spends. We think we are faster as well (our first proof-of-concept study took about 9 months, which is faster but definitely not 1000X as fast), but I think the most important thing is to find something that actually works to treat patients. That's what we want to prove out with backing from YC.
YC doesn't provide advice about this in a systematic way, but we have been trying to learn how to engage the lay public mostly from talking to other YC founders. It's hard to communicate the exact details that we want to share, but I hope that the spirit of our ideas comes across clearly.
We have been actively validating targets, and in fact Alice's PhD work was centered on performing the in vitro and in vivo validation experiments for computational drug leads. We aren't asking for people to believe in magic, and ultimately we hope that our preclinical experiments will reveal whether or not we have something.
I really appreciate your feedback and of course we have studied a lot of what worked (and mostly, didn't work) in systems biology over the past decade, but the field has shifted dramatically since 2004. A lot more data is available, and experiments are designed much more carefully nowadays. In fact, the thing we worry about most is that the field still isn't rigorous enough and the quality of the data still isn't good enough to translate into druggable targets. And we don't intend to belittle the work of scientists who really laid the foundation for what we are trying to do today, but surprisingly few of the big pharma companies we talked to have active systems biology programs for neurological diseases (we have also learned about others, like at Janssen, that are just getting started). I think it's not a question of if systems biology will lead to new therapies, but rather when the methods and field are mature enough.
Also regarding articles, please email me at [email protected] and I will send you PDF versions to read.
While cheaper sequencing hasn't really produced much in the way of new treatments (yet), I disagree that it hasn't been productive or that sequence data is of low accuracy. For example, it is striking to see how relatively little was known about Mendelian disease genes before 1990 (see https://www.genome.gov/Multimedia/Slides/GSPFuture2014/04_Mc..., slide 6, as an example for retinal diseases). I think one of the most frustrating things now is that the price of sequencing isn't continuing to decrease.
You bring up another good point about "Eroom's Law" in which pharma productivity is actually decreasing, and that article was actually one of the things that got us interested in trying to do something new. What we are trying to do is to help reverse that trend by doing something different than what large pharma companies are already working on - if we keep doing the same thing despite ever-decreasing productivity, then I don't see how we can expect the trend to change.