Hello Bioinformatics lovers,

It is August 8th, the date I came to the States 17 years ago.

Hard to believe that I have stayed about the same years in the US as in my hometown in China.

I became a bioinformatician many years after I came here.

Listen to my story in this new podcast, "Crashing Excel to Curing Cancer — How AI and Open Science Are Rewiring Drug Discovery"​

We will talk about experimental design today.

They call us bioinformagicians.

But here’s the truth: We can’t save an experiment that was broken before it began.

You send me RNA-seq data.

Parental cells.
Drug-treated cells.
Resistant cells.

One sample each. No replicates.
And then you ask: “What’s the insight?”

Here’s the answer: there isn’t one.

Without replicates, we can’t estimate variability.

And without variability, we can’t do statistics.

That shiny volcano plot you’re hoping for? It would be fiction.

DESeq2, edgeR, limma—all the tools you’ve heard of—depend on replicates to calculate dispersion.

Without them, there’s no way to tell real biological signal from a pipetting error.

Your “top hit” might be a true discovery… or a ghost. You’ll never know.

We don’t need twenty replicates. But we do need some.

Even two or three per condition gives us something to work with.

Why? Because biology is messy, noisy, and full of surprises.

Studying drug resistance? Add time points. Add replicates.

Account for heterogeneity. Every step matters.

Too many papers are built on shaky ground: no replicates, no controls, no transparency.

Don’t be that lab.

Bioinformatics isn’t magic. It’s a conversation between biology and data. Garbage in = garbage out.

Talk to your bioinformatician before you even start.

A 30-minute conversation can save a 3-month project from ending in nothing.

Key takeaways:

• No replicates = no statistical confidence
• Design matters more than fancy analysis
• Communicate early and often

Action item:​
If you’re planning an experiment, call your data analyst before you touch a pipette.

Science deserves better planning.

Other posts that you may find helpful

1. you can learn anything if you want: ​Open Source Society University.​
2. Depending on your data type, use the right plot to tell the story.
3. One of the biggest hidden killers in bioinformatics?
4. Best Practice for R :: Cheat Sheet​
5. ​Filling Missing Gene Names (NAs) in Genomics Data with {tidyr}
6. "Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines"
7. ​Heatmap in ggplot2 ​
8. ​problems with public scRNAseq datasets.
9. Tired of Venn diagrams? Upset plots are the next-level visualization for intersecting gene sets. Here’s a step-by-step guide to creating them in R. 🧵👇
10. Being a bioinformatician is like being a detective.
11. ​Exploratory Data Analysis (EDA) is the first step in any data analysis journey
12. Linux file permissions explained.​
13. why p-value histogram is useful​
14. ​How to get fasta sequences based on a bed genomics coordinates file​
15. Do not want to debug, use Bash strict mode
16. How to preprocess GEO bulk RNAseq fastq file with salmon ​
17. Using Human Protein Atlas to Find Tissue Specific Genes in R using bioconductor ​

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!