Writing
Technical posts and research notes on AIVC, single-cell biology, and computational herbology.
When TME-VI declines to make a prediction, it's not failing — it's communicating genuine biological ambiguity. Here's what high-entropy cell neighborhoods reveal about immune escape.
Extending a melanoma-trained model to NSCLC via CELLxGENE Census — what transfers, what doesn't, and what that tells us about the universality of immune cell representations.
How do you turn a medicinal herb into a Virtual Cell experiment? A walkthrough of the full pipeline from LC-MS profiling to LINCS L1000 mapping to perturbation prediction.
Vector search retrieves relevant chunks. Knowledge graphs answer relational questions. Here's why CellScout uses both, and when each one wins.
CellChatDB and NicheNet encode known signaling relationships. Can the same prior knowledge infrastructure validate traditional herbal hypotheses computationally?