8/ Science already takes time, I hope to help make it worth it.

Finally, terrified, we sent it to PNAS @pnas.org.
After one round of review, reports came back: supportive. Positive.
Accepted 🎉 🎉

1/ Everything that could go wrong in paper publishing… did.
A story of patience, absurdity, and persistence 🌀 <1min

From Alfonso Valencia’s lab and a very stubborn PhD student (me).

6/ Beyond known links, our approach points to potentially underdiagnosed comorbidities with strong molecular evidence and therapeutic implications.

We made it all explorable here 👉 disease-perception.bsc.es/rgenexcom/

5/ We propose mechanistic models for disease co-occurrences 🔎

- Immune dysregulation was the most common thread
- Specific genes & pathways highlight therapeutic implications
- Even inverse comorbidities emerge (Huntington’s vs cancers, with opposite transcriptional programs)

4/ But patients aren’t homogeneous.
So we stratified them into subgroups with similar gene expression profiles.

This boosted recall to 64% with consistent precision.
And crucially, it revealed numerous subgroup-specific comorbidities.

3/ We built a disease network from large-scale RNA-seq data on human diseases.

👉 It significantly captures ~50% of known comorbidities, far outperforming previous attempts.

🚨 New in PNAS!
🧬 64% of disease co-occurrences can be explained by transcriptomic similarities.

Comorbidities aren’t random—they have a molecular basis.

Here’s how we found it 👇 (1/n)

🔗 doi.org/10.1073/pnas.2421060122

@alfonsovalencia.bsky.social