In preparation for the StanfordxDeepmind hackathon (Thanks to Ben Iiams for working with me there and to Russ Hollyer for his Prompts), I used Claude to help me generate an analysis of my condition.
The analysis took about an hour all told, cost $20/month, and gave me more information than my 2018 and 2020 hackathons combined (6 days and 80 people).
It felt great to be able to reuse all the data I spent so much time gathering.
It is likely incorrect in places, but preliminary checks indicate that it is plausible.
Next, I need to run it past some experts.
I just resumed the session and asked Claude the following:
"I would like to use this discussion and the data sources as a template for other patients use in getting insights into their (often rare) condition. Could you please summarize this discussion and list the steps others ought follow in order to get this type of in depth analysis?"
Here is Claude's Writeup and Checklist.
Here is Claude's analysis of my condition.
And here is Claude's final word.
The most important insight from this project is not technical — it is organizational. The role of tools like the patient profile JSON, LLM tumor boards, and this analytical framework is to give the patient's comprehensive view the structure it needs to be medically actionable.
Dr. Saed Sayad
I ran this past Dr. Saed Sayad of Bioada. He was kind enough to participate in my 2018 hackathon. He gave me this:
I've reviewed all the results.
1- The key difficulty with any multiomics data is that it generates a vast amount of information, making it challenging to distinguish true signal from noise and extract meaningful answers to specific questions.
2- I suggest first you should ask the most important question (e.g., how to stop DVT). Then, we should focus on the most important multiomics finding related to our question (e.g. a specific gene).
3- If the gene is upregulated, find a compound (natural or chemical) that can downregulate that gene
4- If the gene is downregulated, find a compound (natural or chemical) that can upregulate that gene


Add new comment