Nice! Something I'd add about the discovery process is that the safety of tnf-alpha inhibitors was not obvious and is still not well understood. Some big part of why they were hard to develop is that it's weird that blocking tnf-alpha can be effective in so many autoimmune diseases but doesn't seem to cause too much cancer or serious immunosuppression. I think a lot of effort was spent establishing that in animals. Also, this paper (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1766582/) suggests that they were really focused on a fully humanized antibody to have clearly better pk than the existing mabs, which sounds like it was pretty tricky. It reminds me of all the engineering that went into fixing semaglutide's pk.
Interesting comments regarding the importance (or lack thereof) of clinical trials as a rate-limiting step. However the cost of trials is rising inexorably (IQVIA), and the entire clinical process is arguably an oligopoly run by a small number of large CROs. This suggests a sector in need of real disruption. I’d also add that Abbvie is a pretty hard-nosed organisation that was determined to extract every $ of value. Consequently Humira enjoyed even greater success than expected. Thoughts on Eroom’s Law are certainly thought provoking although it’s clear far too early (in my opinion) to suggest that AI directed drug design has really moved the needle regarding POS.
Regarding trials, I want to re-emphasize that prior to writing this, I perceived trials as the primary bottleneck/time constraints on therapeutic development timelines. So when I say learning about Humira updated my view, I mean just that. I still view trials as a major bottleneck, and absolutely agree we should work to make them faster/cheaper/better. But in this case, my takeaway was that they constituted less of the total time than I went in expecting.
Regarding Eroom's and AI, I intentionally didn't mention AI's impact on Eroom's here because I agree it's early to assess. The examples I mentioned, GLP-1s and gene therapy are, to date, unrelated to AI. That said, I suspect I may be more optimistic in AI's likely impact than many coming from the traditional therapeutics world (unsurprisingly given my work/background).
Nice! Something I'd add about the discovery process is that the safety of tnf-alpha inhibitors was not obvious and is still not well understood. Some big part of why they were hard to develop is that it's weird that blocking tnf-alpha can be effective in so many autoimmune diseases but doesn't seem to cause too much cancer or serious immunosuppression. I think a lot of effort was spent establishing that in animals. Also, this paper (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1766582/) suggests that they were really focused on a fully humanized antibody to have clearly better pk than the existing mabs, which sounds like it was pretty tricky. It reminds me of all the engineering that went into fixing semaglutide's pk.
Thanks Trevor, super helpful context. May try to weave this into the timeline/post with a credited edit.
Interesting comments regarding the importance (or lack thereof) of clinical trials as a rate-limiting step. However the cost of trials is rising inexorably (IQVIA), and the entire clinical process is arguably an oligopoly run by a small number of large CROs. This suggests a sector in need of real disruption. I’d also add that Abbvie is a pretty hard-nosed organisation that was determined to extract every $ of value. Consequently Humira enjoyed even greater success than expected. Thoughts on Eroom’s Law are certainly thought provoking although it’s clear far too early (in my opinion) to suggest that AI directed drug design has really moved the needle regarding POS.
Thanks for your comment!
Regarding trials, I want to re-emphasize that prior to writing this, I perceived trials as the primary bottleneck/time constraints on therapeutic development timelines. So when I say learning about Humira updated my view, I mean just that. I still view trials as a major bottleneck, and absolutely agree we should work to make them faster/cheaper/better. But in this case, my takeaway was that they constituted less of the total time than I went in expecting.
Regarding Eroom's and AI, I intentionally didn't mention AI's impact on Eroom's here because I agree it's early to assess. The examples I mentioned, GLP-1s and gene therapy are, to date, unrelated to AI. That said, I suspect I may be more optimistic in AI's likely impact than many coming from the traditional therapeutics world (unsurprisingly given my work/background).