I wish I’d known of this organization, this program, in 2023. My wife was originally misdiagnosed with a very large pulmonary embolism in the major artery connecting the left lung to the heart, very near the heart. The original diagnosis was reasonable, in keeping with the old adage about hoofbeats, horses and zebras.
Only after the PE didn’t respond to blood thinners (and, frankly, after continuous pressure from us to consider a differential diagnosis), she was diagnosed with Pulmonary Artery Sarcoma.
A pneumonectomy and partial resection of the artery, followed by aggressive chemotherapy and radiation therapy, bought her some time. I’m told, and I accept, that this ultra rare cancer was not curable (the cancer, after all, is in the artery wall, inside the circulation system and hence beyond any possibility of localization, isolation), but perhaps there’s some obscure match in this database with a treatment which could have bought even a bit more time.
I’d be interested to see more study of this system, its successes, strengths and weaknesses, including success rates when there is an apparent match.
This is one of those posts that leaves you both inspired and a little unsettled, in the best way.
David Fajgenbaum’s story is extraordinary on the human level (a med student banking his own blood, doing experiments between hospitalizations, finding a decades-old transplant drug that no one had tried, and then living 12 years in remission), but the deeper lesson is systemic: we don’t just have a discovery problem, but we have a connection problem. 
The angiosarcoma example you mention is the part that should haunt all of us: a 2013 signal (high PD-L1 expression), three years of delay before anyone acted, and then a patient with a decade-long remission once the dots were connected. That gap, between what’s already in the literature and what reaches a real patient in time, is where so much preventable suffering lives. 
I love the ambition of Every Cure’s framing: scoring ~4,000 approved drugs against ~18,000 diseases to systematically surface “on-the-shelf” candidates. AI may not be the cure, but it can absolutely be the flashlight. The hard work after the flashlight is incentives, trial infrastructure, and rapid, ethical pathways to test repurposed hypotheses, especially in rare disease where the alternative is often “nothing.”
I wish I’d known of this organization, this program, in 2023. My wife was originally misdiagnosed with a very large pulmonary embolism in the major artery connecting the left lung to the heart, very near the heart. The original diagnosis was reasonable, in keeping with the old adage about hoofbeats, horses and zebras.
Only after the PE didn’t respond to blood thinners (and, frankly, after continuous pressure from us to consider a differential diagnosis), she was diagnosed with Pulmonary Artery Sarcoma.
A pneumonectomy and partial resection of the artery, followed by aggressive chemotherapy and radiation therapy, bought her some time. I’m told, and I accept, that this ultra rare cancer was not curable (the cancer, after all, is in the artery wall, inside the circulation system and hence beyond any possibility of localization, isolation), but perhaps there’s some obscure match in this database with a treatment which could have bought even a bit more time.
I’d be interested to see more study of this system, its successes, strengths and weaknesses, including success rates when there is an apparent match.
This is incredible. It's wonderful to hear more about the positive side of AI when it seems scary much of the time.
This is one of those posts that leaves you both inspired and a little unsettled, in the best way.
David Fajgenbaum’s story is extraordinary on the human level (a med student banking his own blood, doing experiments between hospitalizations, finding a decades-old transplant drug that no one had tried, and then living 12 years in remission), but the deeper lesson is systemic: we don’t just have a discovery problem, but we have a connection problem. 
The angiosarcoma example you mention is the part that should haunt all of us: a 2013 signal (high PD-L1 expression), three years of delay before anyone acted, and then a patient with a decade-long remission once the dots were connected. That gap, between what’s already in the literature and what reaches a real patient in time, is where so much preventable suffering lives. 
I love the ambition of Every Cure’s framing: scoring ~4,000 approved drugs against ~18,000 diseases to systematically surface “on-the-shelf” candidates. AI may not be the cure, but it can absolutely be the flashlight. The hard work after the flashlight is incentives, trial infrastructure, and rapid, ethical pathways to test repurposed hypotheses, especially in rare disease where the alternative is often “nothing.”