Hi Bioinformatics lovers,

My friend Kevin Libuit asked AI to diagnose his broken water heater. It sent him to check one part, then another, then a third.

He still didn’t know what was wrong, so he called a plumber.

Kevin told me this story on his new podcast Down to a science, and it describes what I see every day using Claude Code for bioinformatics.

AI multiplies what you already know. If you know a lot, you move fast.

If you know little, you get confident answers you can’t evaluate.

The question you didn’t know to ask

Ask AI to write a pipeline that aligns reads and builds a count matrix. It will.

But do the default parameters fit your data?

Do they suit your read length?

If you have never built that pipeline yourself, you won’t think to ask, and the AI won’t flag it. It answers every question, including the wrong ones, with the same confidence.

I spent 14 years making mistakes, debugging programs that wouldn’t run, and Googling my way out.

That struggle built the judgment I bring to every AI session now.

Where I use AI, and where I don’t

Deterministic tasks get deterministic code. FASTQ to VCF and FASTQ to counts run through Nextflow pipelines our team has tested for years.

I can trace every parameter, and the same input gives the same output.

AI earns its place at judgment calls. Is this low-quality sample still usable for this experiment?

Does this downstream analysis make sense for this target?

I use AI as a thinking partner, then check its ideas with the wet-lab biologist who has studied that target for a decade.

One workflow I rely on: AI writes the R or Python script, and I commit it to Git. When a wet-lab scientist needs to run it, I wrap the script in a Shiny app on our internal server.

They upload a spreadsheet and get reproducible results, because the app runs the same validated script every time.

If you’re early in your career

Use AI as a tutor. Ask it for a learning plan for Unix, R, or statistics.

Ask it to test your understanding. Learn the terminal anyway.

When the code breaks, you need to know why.

Where AI will matter most in drug development

Kevin’s last question: where will AI make the biggest difference? My answer was clinical trial operations.

A phase 3 trial can take three years from first patient to data close.

Finding patients, enrolling them, and coordinating logistics slow everything down.

Cut one year, and a blockbuster drug gains a year of sales worth a billion dollars or more.

Discovering new biology with AI stays hard because the data is sparse.

Speeding up the process is within reach. Linkedin post https://www.linkedin.com/feed/update/urn:li:share:7520574838801973248/​

​

Where do you draw the line between AI and deterministic code in your work? Hit reply and tell me.

Happy Learning!

Tommy aka crazyhottommy

PS:

If you want to learn Bioinformatics, there are four ways that I can help:

1. My free YouTube Chatomics channel, make sure you subscribe to it.
2. I have many resources collected on my github here.
3. I have been writing blog posts for over 10 years https://divingintogeneticsandgenomics.com/​
4. follow me for daily bioinformatics dose https://www.linkedin.com/in/%F0%9F%8E%AF-ming-tommy-tang-40650014/​

Stay awesome!