Fran Supek

Figure 1: Data modalities of genomic language models. From review article "The DNA dialect: a comprehensive guide to pretrained genomic language models" by Veiner and Supek.

On the heels of the release of #AlphaGenome, we have written up a review on 🧬genomic language models for DNA/RNA, out in EMBO @molsystbiol.org journal

🔖"The DNA dialect: a comprehensive guide to pretrained genomic language models" by Marcell Veiner👏 & yours truly👇

link.springer.com/article/10.1...

We also have a PhD student position open. Application deadline 10.2. 👇
candidate.hr-manager.net/ApplicationI...

Postdoc in AI for Knowledge Graphs in Genomics We are looking for a highly motivated and dynamic postdoc for a 3-year position, to commence 15th April 2026, or soon thereafter (flexible start date). The PhD

Ideal positions for recent bioinformatics/genomics/compbiol PhDs looking to pivot towards AI 🖥️ 🧠 & to do impactful science. Also great for computer science people looking to get into genomics. 🧬

1️⃣ candidate.hr-manager.net/ApplicationI...

2️⃣ candidate.hr-manager.net/ApplicationI...

We're recruiting 2 postdocs👩‍💻👨‍🏫 on the "A-SOuRCCE" project:

1️⃣AI for Knowledge Graphs in Regulatory Genomics (auto extraction, organization and querying of knowledge)
2️⃣Agentic AI for Scientific Discovery in Genomics (LLMs for hypothesis generation and assessment)

Apply below! 👇 Deadline soon: 10.2.

AI-generated cartoon illustrating a team of researchers working on A-SOuRCCE

Big thanks to my awesome lab members 💯 @genomedatalab.bsky.social, and to present & past host institutions who helped shape this vision @bric-ucph.bsky.social 🥳 @irbbarcelona.org 🙌 @crg.eu 🎉@institutrb.bsky.social🙏

Now let’s get to work!🔬🤖 Will be recruiting PhDs/postdocs soon, watch this space...

By integrating AI agents with knowledge graphs grounded in biological fact, our discovery engine will:

- auto-generate mechanistic hypotheses 🧬
- assess their novelty 💫 and plausibility 🧠
- validate against population genetics (GWAS) data 📊

This will accelerate pathway: data -> mechanism -> target

AI-generated cartoon to illustrate the central idea of the A-SOuRCCE project

Delighted to share that our project on agentic AI for genomic science has been funded!🌟 We secured 1.7M € from @novo-nordisk.bsky.social in “Data Science Investigator: Distinguished” - v grateful!🙏

A-SOuRCCE stands for "AI for Single-cell Omics and Reproducible Cardiometabolic & Cancer Exploration"

Intrinsic DNA sequence determinants and tissue-specific regulation of human replication origins The accurate duplication of the genome relies on the spatiotemporal control of DNA replication initiation, yet the determinants specifying mammalian origin locations remain elusive. By developing ORIF...

Overall: human DNA replication origins usage isn't a roll of the dice 🎲 it’s genetic potential gated by epigenetic context

👏Huge congrats to lead authors Marcell Veiner
& Marina Salvadores - bravo for excellent teamwork!

Read the full study here...👇
www.biorxiv.org/content/10.6...

2) the "software"

🩸🫁DNA is same in every cell - why do colon/lung/... cells copy differently?

🦀We tracked "fossil" mutations in cancer genomes to decode tissue-specificity

🧬Local chromatin environment acts as a switch, turning the hard-coded sites ON or OFF in each tissue

1) the "hardware"

💻🧠 We trained a deep learning model (ORIFormer), revealing replication initation is "hard-coded" in DNA by specific sequence motifs, like ZNF770 sites

🧬 The genome has a built-in map of potential start sites, w/ complex "grammar" that was previously elusive

AI-generated graphical summary of the DNA replication origins study

📢new study @genomedatalab.bsky.social

When a cell divides, it copies 3 billion DNA letters. How does it know where to begin?

⁉️For years, the field debated if replication initiation was random or strict

📜Our preprint uses AI + mutation signatures to show it’s a precise, two-part system

Dual inhibition of the nonsense mediated mRNA decay enhances tumour immunogenicity, drives immunoediting, and potentiates checkpoint blockade Immune checkpoint blockade has transformed cancer therapy, but current biomarkers such as tumour mutation burden often fail to reliably predict clinical benefit. One proposed reason for this discrepan...

👉Read the study by Zadra, Orsolic et al.

"Dual inhibition of the nonsense mediated mRNA decay enhances tumour immunogenicity, drives immunoediting, and potentiates checkpoint blockade"

as a biorxiv preprint, fresh from the oven:

www.biorxiv.org/content/10.6...

AI-generated cartoon displaying simplified concept

➡️ We trained protein language model on these datasets of mutations from tumors treated by NMDi. A fine-tuned 💻AI model successfully learned to identify immunogenic neopeptides, capturing features that go beyond standard MHC-binding predictions.

AI-generated cartoon displaying simplified concept

➡️ Our mRNA data adds to the mounting literature that abberant splicing in cancer (esp intron retention) could contribute to making tumors immune-hot. 💊NMD inhibiton is extremely helpful in exposing these tumor splicing errors.

AI-generated cartoon displaying simplified concept

➡️ Using WGS, we tracked the evolutionary footprint of treatment. We observed strong "immunoediting" -- a negative selective pressure against immunogenic mutations -- thus NMD inhibition makes tumors highly visible to the immune system.

Dual inhibition of the nonsense mediated mRNA decay enhances tumour immunogenicity, drives immunoediting, and potentiates checkpoint blockade Immune checkpoint blockade has transformed cancer therapy, but current biomarkers such as tumour mutation burden often fail to reliably predict clinical benefit. One proposed reason for this discrepan...

Our team's contribution @irbbarcelona.org (led by the mighty postdoc 🤓 Davor Orsolic) and yours truly at Biotech Research & Innovation Centre (BRIC), UCPH_Research dived into the 🧬 tumor genomics and computational modeling.

www.biorxiv.org/content/10.6...

read more!👇

In a synergy with the @TheAnaJanic lab @UPFBarcelona, who led the study spearheaded by Ivan Zadra 👏, we investigated how inhibiting the cellular quality-control mechanism known as NMD can force lung tumors in 🐁 to reveal their hidden mutations to the immune system.

Figure summarizing NMD from a review article Supek, Lehner and Lindeboom (2021) Trends in Genetics.

Excited to share our latest collaborative study of a frontier approach in cancer immunotherapy, targeting the Nonsense-Mediated mRNA Decay (NMD) pathway.

Read more in a short 🧵+ link to preprint is below:

bric.ku.dk

candidate.hr-manager.net/ApplicationI...

📣 Attention postdocs in the life sciences! Are you ready to take the big leap?

Start your lab at the Biotech Research and Innovation Centre in Copenhagen! 🇩🇰 @bric-ucph.bsky.social

Collegial atmosphere, cutting-edge science, core funding.

(+the weather is wonderful, relative to Siberia😁)

Apply!👇

Congrats to colleagues from IRB!

Variable efficiency of nonsense-mediated mRNA decay across human tissues, tumors and individuals - Genome Biology Background Nonsense-mediated mRNA decay (NMD) is a quality-control pathway that degrades mRNA bearing premature termination codons (PTCs) resulting from mutation or mis-splicing, and that additionally participates in gene regulation of unmutated transcripts. While NMD activity is known to differ between examples of PTCs, it is less well studied if human tissues differ in NMD activity, or if individuals differ. Results We analyzed exomes and matched transcriptomes from Human tumors and healthy tissues to quantify individual-level NMD efficiency, and assess its variability between tissues, tumors, and individuals. This was done by monitoring mRNA levels of endogenous NMD target transcripts, and additionally supported by allele-specific expression of germline PTCs. Nervous system and reproductive system tissues have lower NMD efficiency than other tissues, such as the digestive tract. Next, there is systematic inter-individual variability in NMD efficiency, and we identify two underlying mechanisms. First, somatic copy number alterations can robustly associate with NMD efficiency, prominently the commonly-occurring gain at chromosome 1q that encompasses two core NMD genes: SMG5 and SMG7 and additional functionally interacting genes such as PMF1 and GON4L. Second, deleterious germline variants in genes such as the KDM6B chromatin modifier can associate with higher or lower NMD efficiency in individuals. Variable NMD efficiency modulates positive selection upon somatic nonsense mutations in tumor suppressor genes, and is associated with cancer patient survival and immunotherapy responses. Conclusions NMD efficiency is variable across human tissues, and it is additionally variable across individuals and tumors thereof due to germline and somatic genetic alterations.

Check out the full story here:

"Variable efficiency of nonsense-mediated mRNA decay across human tissues, tumors and individuals"

genomebiology.biomedcentral.com/articles/10....

👏👏to Guille for marvelous work!

To NMD or Not To NMD: Nonsense-Mediated mRNA Decay in Cancer and Other Genetic Diseases The nonsense-mediated mRNA decay (NMD) pathway degrades some but not all mRNAs bearing premature termination codons (PTCs). Decades of work have elucidated the molecular mechanisms of NMD. More recent...

Why care? Our prior work (Lindeboom et al., reviewed in www.cell.com/trends/genet... ) + Guille paper supports that NMD

🛑modulates selection on somatic nonsense "driver" mutations
🛑associates w/cancer (immuno)therapy💊response (neoantigens!)
🛑linked w/ severity of inherited genetic disease

👇

A key genetic safeguard works differently across tissues and individuals A puzzling question in molecular biology has been why our bodies do not uniformly eliminate faulty genetic information—especially that which could lead to disease. A key mechanism responsible for this...

Interestingly, genetic variants can change global NMD efficiency in an individual🤔

🧬 inherited variants - SNVs in KDM6B chromatin modifier gene
🧬 somatic variants - copy number alterations at chromosome 1q (SMG5, SMG7...)

Nice IRB writeup here 👇
www.irbbarcelona.org/en/news/scie...

🛑Guille's @guillepalou.bsky.social no-nonsense 😉 paper is out in @springernature.com Genome Biology!

We report variable activity of the NMD pathway (mRNA QC) across tissues & individuals:

🧠less active in the nervous system & in reproductive tissues
🩸more active in digestive tract, muscle, blood

Full dataset here
👇
elsevier.digitalcommonsdata.com/datasets/btc...

Fantastic to be in such great company!🤩

👏Congrats to 19 colleagues from my institutions (🇩🇰BRIC @UCPH_health and 🇪🇸@IRBBarcelona) in this year's Elsevier/Stanford top 2% most cited researchers list

Thanks to organizers for inviting!

🧬New approach to decoding mutational signatures improves prediction of prognosis in #OvarianCancer.

📰 #GenomeMedicine ✍️ Patricia Ferrer-T Iván Galván-F & @fransupek.bsky.social

➡️ bit.ly/4nA579b 📌DOI: 10.1186/s13073-025-01497-7

@genomedatalab.bsky.social @bric-ucph.bsky.social

🧪

Congrats to our CRISPy fresh doctor 👏👏 Marcel McCullough for successfully defending his PhD thesis:

🧬✂️"Combinatorial Gene Editing in Human Cells to Model Deficiencies of DNA Repair in Cancer"

Thx to committee members Julian Ceron, Laura Valle and David Szuts for a great discussion!