Hello Bioinformatics lovers,

Tommy here. Recursion, the AI star company, recently laid off 20% of its people.

It is a humble reminder that Drug discovery is hard.

Biology is more complicated than many thought.

AI drug companies that do not have deep domain biology knowledge will inevitably fail.

Today's newsletter is a simple example: Why Your CRISPR Screen Hit Might Be a Trap​
​From bioinformatics insight to therapeutic reality.

You ran a CRISPR screen.​
You found top hits that kill cancer cells.
You think: this is it.​
Your next paper. Your next drug.
But then… it unravels.

Your top hits?​
Ribosomal genes.
POLR2A.
Essential for every cell—not just cancer.

You didn’t find a drug target.
You rediscovered how life works.

Here’s the catch:​
CRISPR screens often highlight common essential genes—those every cell needs to survive.

Great for killing cells in vitro.
Disastrous in a living human body.

This is where biology matters.​
A good cancer target should be:

• Essential in cancer cells
• Dispensable in normal tissue
• Selectively druggable

How do you know?​
You look beyond the CRISPR screen:

• GTEx: Gene expression in normal tissues
• Human Protein Atlas: Protein-level data
• TCGA: Tumor vs. normal comparisons

Ask hard questions:​
Your gene is overexpressed in tumors—great.
But is it also high in heart?
Liver?
Kidney?

Because your drug will reach those too.

This is why Antibody-Drug Conjugates (ADCs) matter.

• Find a surface protein highly expressed in cancer
• Build an antibody to it
• Link it to a toxin

Result: Precision delivery that kills the tumor, not the patient.

CRISPR gives you hits.​
But drug development asks deeper questions:

• Can you hit it without hurting everything else?
• Is it even druggable?
• Will it work in vivo?

In vitro is clean.​
​In vivo is chaos.
The bloodstream doesn’t care about your p-values.

Takeaways:​
✔️ CRISPR hits ≠ druggable targets
✔️ Check expression in normal tissues (GTEx, HPA)
✔️ Specificity > potency
✔️ Safety before celebration

Final thought:​
Bioinformatics is not just about small p-values.
It’s about bridging discovery and therapy.

You’re not just running pipelines.
You’re choosing what’s worth chasing.

And that requires science—and a whole lot of judgment.

Other posts that you may find helpful:

1. ​why you should learn linear algebra well for bioinformatics.
2. Want to master bioinformatics (or any skill)? Here's the hard truth: There’s no shortcut to success.
3. I used to spend a week curating single-cell studies by hand. Now AI does it in 5 minutes.​
4. Are you ready to level up your bioinformatics skills? Let’s talk about repetition—a key concept that can save you hours in real-world data analysis.
5. Here’s my 3-step framework to learn Bioinformatics or learn anything in life 👇
6. What is k-means clustering, and how can you make its results consistent? Let’s break it down.
7. Ever feel buried under 20,000 genes and no clue where to start? That’s the curse of dimensionality in RNA‐seq
8. How to read GB-size file into R fast!
9. ​Machine learning isn’t a flowchart. it is an art.​
10. Here is my story from wet lab scientist to a computational biologist again.
11. what is scRNAseq sparse matrix​

Happy Learning!

Tommy aka, crazyhottommy

PS:

If you want to learn Bioinformatics, there are other 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/​

​

Stay awesome!