Dear Bioinformatics lovers,

Looking at bioinformaticians’ profiles today, it seems like everyone is immersed in cutting-edge single-cell analyses and AI-driven bioinformatics.

It’s inspiring to see how far the field has come, but amidst all the buzz, something crucial is often overlooked—the fundamentals of bioinformatics.

Why Fundamentals Still Matter

Basic bioinformatics skills, like exploratory data analysis (EDA) and data sanity checks, seem to be losing the spotlight.

Everyone’s talking about deep learning pipelines and generative AI models, but these advanced techniques often depend on a solid foundation in the basics.

Bioinformatics isn’t just about fancy models. It’s about understanding messy, real-world data—quality control, normalization, and ensuring the data makes sense before diving into machine learning.

The pioneers of this field didn’t start with deep learning on single-cell data or transformer models for genomics.

They began by asking simple, meaningful questions, doing EDA, and spotting biological patterns.

This approach has led to some of the most impactful breakthroughs.

Unfortunately, it’s becoming harder to find bioinformaticians who take the time to question their data and ask, “Does this make sense biologically?” Too often, there’s a rush to fit the latest model without fully understanding the data.

Sure, learning machine learning is valuable—I’m exploring the fast.ai course myself—but the real power in bioinformatics lies in knowing when to trust your data and when to question it.

My advice? Before chasing the latest trend, master the basics.

Learn how to clean, explore, and deeply understand data. These skills will take you further than any buzzword ever will.

Connecting Biology to Bioinformatics

Building on this foundation, understanding the biology behind your data is equally essential.

Without it, even the most rigorous analyses can lead to flawed conclusions.

One critical example is the discrepancy between RNA and protein levels. Relying on RNA-seq data alone might not tell the whole story.

Proteins are regulated at multiple levels, and this can lead to significant differences in what RNA and protein levels reveal about a system.

Take CTLA-4 in T cells:

• CTLA-4 is an immune checkpoint protein.
• Its mRNA may be present, but the protein’s surface expression is tightly controlled by intracellular trafficking and recycling.
• Without understanding this, you might misinterpret immune cell activity.

Or consider IFNAR1 after interferon-alpha stimulation:

• While interferon receptor mRNA decreases slightly, protein levels drop dramatically due to ubiquitination and degradation.
• Analyzing mRNA data alone might lead you to conclude minimal changes, missing the actual biological impact.

Another classic example is HIF-1α under hypoxia (my first PhD paper is on how CTCF blocks HIF1 enhancers :):

• HIF-1α mRNA remains steady under normoxia, but its protein is degraded unless hypoxia stabilizes it.
• Without context, you might overlook hypoxia’s critical role in gene regulation.

Multi-Omics Is the Key

To address these gaps, integrating RNA and protein data—multi-omics—offers a more complete picture:

• You can validate findings across modalities.
• Post-transcriptional regulation becomes apparent.
• Complex biological systems become easier to decipher.

In single-cell studies, RNA and protein often tell complementary stories. For example:

• CD4 T cells might show low mRNA levels for CD4 but high protein expression.
• NK cells often have low CD56 mRNA but high protein levels.

If you rely solely on RNA data, you’ll miss crucial phenotypic details.

Even if you have a high protein level, their activity may be regulated by post-translational modifcation (e.g., phosphorylation)

Takeaways

1. Master the basics: Start with EDA, data cleaning, and normalization. These skills will always be relevant, no matter how advanced the field becomes.
2. Understand the biology: Biological knowledge provides the context needed to interpret your data accurately.
3. Adopt multi-omics approaches: Combining data modalities like RNA and protein enables more robust and insightful analyses.

By focusing on these principles, you’ll not only become a better bioinformatician but also a more impactful scientist.

Let me leave you with a question: What are your favorite examples of RNA-protein mismatches? I’d love to hear your thoughts!

Happy Learning!

Tommy aka, Crazyhottommy
​

PS: my other Linkedin posts that you find find helpful for the past week:

1. ​Learning Linux for bioinformatics? Understand file paths first!​
2. ​Need to copy files to your HPC or cloud machine? Meet your new best friend: rsync. Here’s why it’s better than scp and how to use it like a pro.​
3. ​What the heck is an object (Seurat object?) in object-oriented programming? 🤔 Let's break it down with bioinformatics examples
4. The most underrated Unix command for bioinformatics: ln -s (soft link)
   ​Here’s why it’s so powerful, how it works, and how I use it to simplify my life in bioinformatics
5. Why I love Unix tools for bioinformatics—and why you should turn your R scripts into command-line tools ​
6. Bioinformaticians: Tired of losing your remote sessions on HPC or cloud VM? Let me introduce you to tools that will save you time: screen, tmux, and mosh.
7. ​What’s the most influential course I’ve taken for computational genomics?​
8. ​When NOT to use DESeq2 for RNA-seq analysis?​
9. Machine learning (ML) is revolutionizing genomics, but common pitfalls can lead to misleading results. Here's a thread on how to avoid them.

PPS:

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!