Third day doing Bioinformatics
Checking in late today.
I switched now to the Colab Pro sub for 11 €, which in my opinion gives me the most GPU time for the least amount of money. Unfortunately the runtime also tends to disconnect there after a few hours, so I had to redo the run i did overnight and added meticulous logging of checkpoints.
I was mostly running cells and copy-pasting log messages and errors to Claude Code and now i was able, at least for the lung tumor dataset, to show that you dont have to another costly hyperparameter search.
But I need to find other datasets to check also for them if this finding holds. And that seems to be a big issue. Claude found another dataset that I will look into tomorrow, which contains CITE-seq data from melanoma cells.
I was tired at the end of the day so I dumped everything of today into a huge .md file, pasted it in a new Claude.ai Chat session, opened that in the Claude mobile app and just recapped about the day via Voice mode while laying in bed, eyes closed. I enjoyed that.
I also reworked parts of the website and also want to start filling in the real data into the trials dashboard.