What bioinformatics taught me about asking better questions

Nobody warns you that learning bioinformatics is mostly learning to be precise. A pipeline does not care what you meant. It runs exactly what you wrote, on exactly the data you gave it, and returns exactly what that implies, no more and no less.

Coming from qualitative training, this was a shock. Anthropology rewards nuance, hedging, context. Code rewards none of these. My first scripts failed in ways that were, in hindsight, failures of thinking: undefined samples, unstated assumptions, comparisons I had not actually justified.

The discipline of the pipeline

Over time I noticed the rigor traveling backwards into my anthropological work. Before I run an analysis now, I write the question in one sentence. If I cannot, I am not ready. Some habits that stuck:

• State the population, the comparison, and the expected signal before touching data.

• Treat every parameter choice as a claim that needs defending.

• Keep a lab notebook like an ethnographer keeps field notes: dated, honest, complete.

A two-way street

The exchange runs both ways. Anthropology reminds me that reference panels are not neutral, that categories like population and ancestry are loaded, and that the people whose genomes we study are stakeholders, not samples. The best computational work I know is done by people who carry both sensibilities at once.

That is the researcher I am trying to become, and what I will keep writing about here.