Lightnews — Scholar-powered news

Ming Tommy Tang @tommytang.bsky.social · 41m

I was featured in The Data Wire by Pure Storage.
Data challenge comes before the algorithm/AI challenge.

Ming Tommy Tang @tommytang.bsky.social · 15h

I hope you've found this post helpful.

Follow me for more.

Subscribe to my FREE newsletter chatomics to learn bioinformatics divingintogeneticsandgenomics.ck.page/profile

Hi! I'm Tommy Tang

I am a bioinformatician/computational biologist with six years of wet lab experience and over 12 years of computation experience. I will help you to learn computational skills to tame astronomical data and derive insights. Check out the resources I offer below and sign up for my newsletter!

divingintogeneticsandgenomics.ck.page

Ming Tommy Tang @tommytang.bsky.social · 15h

12/
Want to explore intronic reads or retained introns?
Start here:
pmc.ncbi.nlm.nih.gov/articles/PM...

Let your RNA speak fully—even the introns.

Covering all your bases: incorporating intron signal from RNA-seq data

RNA-seq datasets can contain millions of intron reads per library that are typically removed from downstream analysis. Only reads overlapping annotated exons are considered to be informative since mature mRNA is assumed to be the major component ...

pmc.ncbi.nlm.nih.gov

1 1

Ming Tommy Tang @tommytang.bsky.social · 15h

11/
Key Takeaways
Intronic reads are common

Don't ignore them—understand them

Clean RNA prep helps

IR ≠ error; it can be biology

Always validate with orthogonal methods

1

Ming Tommy Tang @tommytang.bsky.social · 15h

10/
Not Always Noise
Intronic reads ≠ garbage.
They may tell you about:
Transcriptional burst

Splicing kinetics

Retained introns with function

Context matters.

1 1

Ming Tommy Tang @tommytang.bsky.social · 15h

9/
When To Worry
Metric Normal (PolyA) Red Flag
Exonic Reads >60% <50%
Intergenic Reads <10% >20%
Intronic Reads ~5–15% >30%

1

Ming Tommy Tang @tommytang.bsky.social · 15h

8/
Fix It (or At Least Understand It)
Use DNase treatment

Check % exonic/intergenic with RSeQC

Try tools like superintronic for IR analysis github.com/sa-lee/supe...

GitHub - sa-lee/superintronic: A tidy interface for coverage analysis

A tidy interface for coverage analysis. Contribute to sa-lee/superintronic development by creating an account on GitHub.

github.com

1

Ming Tommy Tang @tommytang.bsky.social · 15h

7/
Aligner & Annotation Issues
Even splice-aware tools like STAR and HISAT2 struggle with:
Novel isoforms

Pseudogenes

Incomplete annotations .

1

Ming Tommy Tang @tommytang.bsky.social · 15h

6/
Ribo-depletion Libraries
They pull in everything—including lncRNAs and pre-mRNAs.
That’s great for total transcriptome.
Not great if you expect clean mRNA.

1

Ming Tommy Tang @tommytang.bsky.social · 15h

5/
Library Prep Matters
PolyA vs Total RNA makes a big difference.
Method Exonic Intronic
PolyA 69% 7%
Total RNA 56% 21%

1

Ming Tommy Tang @tommytang.bsky.social · 15h

4/
Biological Intron Retention (IR)
This isn’t noise—it’s regulation.
IR plays a role in cell stress, cancer, and mRNA decay.
e.g., HNRNPD, TP53 .

1 2

Ming Tommy Tang @tommytang.bsky.social · 15h

3/
Genomic DNA Contamination
Bad RNA prep? DNA might still be there.
It mimics intronic reads.
You’ll see odd insert sizes or discordant read pairs .

1

Ming Tommy Tang @tommytang.bsky.social · 15h

2/
Pre-mRNA Contamination
Poly(A) libraries still catch unspliced RNA.
Why? Because splicing isn’t instant.
Correlations: intron vs exon counts often r = 0.7–0.9 .

1

Ming Tommy Tang @tommytang.bsky.social · 15h

1/
Your RNA-seq targets mature mRNA, right?
So why do you see intronic reads?
Here’s why—some technical, some biological.

1

Ming Tommy Tang @tommytang.bsky.social · 15h

Why are there intronic reads in your bulk RNA-seq data?
You're not alone—it's common, and the reasons are more layered than you think.
Let’s break it down. 🧵

1 1 5

Ming Tommy Tang @tommytang.bsky.social · 16h

AI deals show no sign of slowing www.nature.com/articles/d4...

AI deals show no sign of slowing

Biopharma Dealmakers - Over the past year, AI-focused biopharma collaborations have increasingly attracted large upfront payments and extended their reach outside of small molecule drug discovery.

www.nature.com

Ming Tommy Tang @tommytang.bsky.social · 16h

I hope you've found this post helpful.

Follow me for more.

Subscribe to my FREE newsletter chatomics to learn bioinformatics divingintogeneticsandgenomics.ck.page/profile

Hi! I'm Tommy Tang

I am a bioinformatician/computational biologist with six years of wet lab experience and over 12 years of computation experience. I will help you to learn computational skills to tame astronomical data and derive insights. Check out the resources I offer below and sign up for my newsletter!

divingintogeneticsandgenomics.ck.page

Ming Tommy Tang @tommytang.bsky.social · 16h

14/
So next time you see only 5 DE genes…
Or only 100 peaks from your TF of interest…
Don’t give up.
Zoom out.
Ask better questions.
And adjust.
Biology won’t fit in your default settings.
Neither should your analysis.

1

Ming Tommy Tang @tommytang.bsky.social · 16h

13/
Bioinformatics isn’t about being strict.
It’s about being smart.
Knowing when to trust the defaults, and when to question them.

1

Ming Tommy Tang @tommytang.bsky.social · 16h

12/
Always go back to the genome browser.
Visualize your bigwig.
Zoom into the peaks.
Some truths are in the noise.

1

Ming Tommy Tang @tommytang.bsky.social · 16h

11/
Use default thresholds—but don’t worship them.
Ask yourself:
Does this make sense biologically?

What am I missing with this filter?

Can visualization help?

1

Ming Tommy Tang @tommytang.bsky.social · 16h

10/
Key takeaway:
Bioinformatics is not a vending machine.
You don’t just plug in data and get out truth.
It’s a craft.

1

Ming Tommy Tang @tommytang.bsky.social · 16h

9/
Don’t forget QC.
High duplication rate?
Low FRiP score?
High mitochondrial reads?
All affect confidence—but don't dictate a hard stop.
Always combine data quality with domain knowledge.

1

Ming Tommy Tang @tommytang.bsky.social · 16h

8/
This isn’t cherry-picking.
It’s being biology-aware.
A dataset can be messy and still reveal truth—if you know where and how to look.

1

Ming Tommy Tang @tommytang.bsky.social · 16h

7/
Maybe the sample was under-chipped.
Maybe sequencing depth was low.
Or maybe your TF just binds loosely
If the biology says this TF should bind thousands of sites—
You adjust the q-value threshold.
You follow the signal.

1