The GenomeComputer Company
Back to tutorials

Tutorial

Write better genome prompts

Turn broad genome questions into precise prompts that ask for files, genes, variants, evidence, sources, and follow-up questions.

6 minAny AI tool

What you will learn:

You will be able to write prompts that produce specific, inspectable .genome answers instead of generic genetics essays.

A genome prompt should tell the AI what data to use, what question to answer, what evidence to show, and how to separate your file from outside knowledge. If you leave those instructions out, the model may give you a polished answer that is not grounded in your .genome.

The best prompts are specific without being narrow-minded. They give the model a topic, ask it to inspect the bundle, request genes and variants, require sources, and invite follow-up questions.

A strong .genome prompt includes

  • The data source: my .genome bundle.
  • The topic: a gene, variant, pathway, trait, paper, or question.
  • The task: inspect, compare, summarize, explain, or build a report.
  • The evidence requirement: show files, annotations, sources, and confidence where available.
  • The output format: table, report, checklist, next questions, or prompt template.

Weak prompt

Tell me about my sleep genetics.

Better prompt

Using my .genome bundle, inspect sleep-related annotations and variants near genes or loci linked to sleep duration, chronotype, sleep quality, or circadian biology. Separate direct findings in my files from annotation context and broader research. Include source names where available and list the best follow-up questions.

Prompt-writing workflow

  1. Write the broad question in plain language.
  2. Name the topic precisely: gene, variant, trait, paper, or pathway.
  3. Tell the model to inspect your .genome before answering.
  4. Ask for direct file evidence first.
  5. Ask for annotation and research context second.
  6. Ask for uncertainty and follow-up questions.
  7. Save the final version as a reusable template.

Reusable exploration template

Using my .genome bundle, investigate [topic]. First identify relevant files, genes, variants, annotations, and sources in my bundle. Then explain what each finding may mean, what evidence supports it, what remains uncertain, and what follow-up questions I should ask next. Return the answer as a concise exploration report.

How to improve any prompt

  • Replace broad words like health, fitness, or nutrition with a specific mechanism or question.
  • Ask for a table when you want comparison.
  • Ask for a report when you want narrative explanation.
  • Ask for source fields when you want provenance.
  • Ask for follow-up questions when you want exploration rather than a final answer.

A good prompt makes the output easier to check. If the answer cannot be inspected, traced, or improved, the prompt probably needs more structure.