Hello Bioinformatics lovers,

Tommy here. My friend Glen recently sent me a LinkedIn post from someone who teaches intro bioinformatics.

The message to students was blunt: entry-level jobs are gone, Claude is smarter and faster than any entry-level research assistant, and the field is doomed.

I have done bioinformatics for 14 years, and teaching it matters to me more than anything else I do. So that post stuck with me.

Here is where I landed. AI makes coding much faster. It also makes the foundations more important, because your job has shifted from writing the code to judging it.

What AI did for me

I use Claude Code almost every day. With about 8 hours of prompting, I ported BETA (the ChIP-seq and RNA-seq integration tool from Shirley Liu’s lab) from Python 2 to Python 3, published it on PyPI, and set up CI/CD. I did the same for ROSE2, the super-enhancer caller. Runtime on 22K hg38 enhancers dropped from about 9 hours to 12 minutes.

In every case I already knew what the right answer looked like. I had used BETA for years. I knew the shape of a super-enhancer hockey-stick plot. AI multiplied skills I already had.

What AI got wrong

AI also answers confidently when it is wrong.

I watched it merge TCGA tumors with GTEx normals where mismatched gene ID versions became zeros, so genes looked tumor-specific.

I saw it z-score a heatmap along the wrong axis. Each fix took minutes once a person noticed.

If nobody notices, the bad result becomes the answer of record.

Catching these takes the boring basics: biology, Unix, one language you can read line by line, git, statistics, linear algebra, and enough machine learning to recognize overfitting and data leakage.

Test 20,000 genes at p < 0.05 and you expect about 1,000 false positives by chance. An AI-generated DESeq2 script can produce a beautiful volcano plot without knowing your design has a batch effect.

The part I can’t answer yet

One senior person with AI can now cover a lot of ground, and many companies have stopped posting entry-level roles.

But every senior bioinformatician was once a junior who got paid to make mistakes where someone else caught them.

I mixed up 0-based and 1-based coordinates more times than I can count. If nobody hires juniors today, where do the experienced people come from in 10 years?

If you’re a student: keep going. Use AI as a tutor. Ask it to explain code line by line, or to quiz you on what you actually understand. Then go analyze real data.

If you’re senior: share more. Be the person you needed when you started.

The full post covers the Reddit debate, my conversation with Kevin Libuit, and the eight skills I tell every junior to learn: https://divingintogeneticsandgenomics.com/post/is-bioinformatics-doomed-future-of-bioinformatics-ai/​

Hit reply and tell me: if you’re a student, what scares you most about entering the field? If you hire, how are you training juniors now?

Stay 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. Lastly, I post on Linkedin daily https://www.linkedin.com/in/%F0%9F%8E%AF-ming-tommy-tang-40650014/recent-activity/all/ Follow me there.

Stay awesome!