Artificial Intelligence Predicts Colorectal Cancer Risk in Ulcerative Colitis Patients A new AI tool developed by UC San Diego researchers uses clinical data to predict colorectal cancer risk in ulcerative colitis patients, helping clinicians make evidence‑based care decisions and preve...

A nice story about the paper also in @ucsandiego.bsky.social Today today.ucsd.edu/story/artifi... @ucsdmedschool.bsky.social @ucsdhealth.bsky.social

Our paper "Automated phenotyping with AI predicts future advanced neoplasia risk in colitis-associated low-grade dysplasia" out today in Clin Gastro Hep! Led by PhD student Brian Johnson-showcases power of LLMs to pull out relevant risk factors in huge EHR datasets www.cghjournal.org/article/S154...

DBMI is seeking applicants for our 2026 Postdoctoral Fellowships in Biomedical Informatics! Join our NIH NLM–funded fellowship. Eligibility: U.S. citizens, non-citizen nationals, or permanent residents (I-551). Apply: dbmi.ucsd.edu/education/nl...

All our methods using open weight LLMs are publicly available to modify for other problems and explained in detail in our paper and SI: github.com/bdj34/llama....

Large language models for extracting histopathologic diagnoses of colorectal cancer and dysplasia from electronic health records Objective Accurate data resources are essential for impactful medical research, but available structured datasets are often incomplete or inaccurate. Recent advances in open-weight large language mode...

Our new paper is out today! PhD student Brian Johnson tour de force in applying and validating LLMs for phenotyping in massive EHR databases. Opens up many exciting opportunities for disease research using population-scale data #mathonco
bmjopengastro.bmj.com/content/12/1...

thank you for the support Trevor!

thank you Adam!

No worries Suzanne, difficult last name club! haha thank you for the support <3

It's open to the public I just walked right on! 7 mins walk from the convention center :)

Gorgeous week in Edmonton, Alberta for #SMB2025 @smbmathbiology.bsky.social Fred Adler giving us the laughs we all need with his plenary on “Education, Bureaucracy and Corruption”

Next up at #EBECancer25 summer school - hands-on digital pathology workshop with @mihaelaangelova.bsky.social @chandler-gatenbee.bsky.social

here this week learning with all of our students at #EBECancer25 @eventswcs.bsky.social ! having a lot of Sudoku-type fun at our “grammar of somatic evolution” with tutor @afrankell.bsky.social

That’s a wrap for QCC lab representation at #DDW2025! So fun to work with these people everyday. I’m also rocking yellow for osteogenesis imperfecta awareness day each May 6th #WishboneDay 💛

Quite an earthquake today in SD! We're all good, but I loved watching the elephants' reaction to a natural threat at our beloved SD Zoo Safari Park 🐘

Congrats to @cgucci.bsky.social on successfully defending her PhD on pi day! First PhD from our lab, so grateful for her dedication to our team and her supportive thesis committee (not pictured: Ludmil Alexandrov, sorry Ludmil!). Very excited to see all she does next... ☺️

Will the recording be up on youtube after streaming? Have a meeting at same time but would like to watch :)

Our Division at UCSD Medicine is hiring postdocs in biomedical informatics on our T15 grant. Please apply if interested or share with others who might be! @ucsdmedschool.bsky.social @ucsandiego.bsky.social @ucsdhealth.bsky.social

Applications are open for our Evolutionary Biology and Ecology of Cancer summer course at the Wellcome Genome Campus, UK! This is a great opportunity for early career scientists/trainees to get plugged into this field #mathonco @amyboddy.bsky.social @trevorgraham.bsky.social @afrankell.bsky.social

Finally, we are very hopeful that this genetic test could help prevent cancers in very high-risk patients. We are pushing this work forward still, much more to come! Thank you to the fantastic team I had the pleasure working with the past years @trevorgraham.bsky.social et al @cancerresearchuk.org

With independent data from other UK centers, we externally validated this biomarker and found similar risk stratification. As a predictive biomarker for HGD/CRC, both PPV & NPV >90% at 5 years post initial LGD resection time point in the validation. Check out the paper for more!

We found LGD from progressors already harbored significantly more CNAs than non-progressors at index Dx. As a binary biomarker, the odds ratio of developing HGD/CRC for a high CNA score was 36, and this was 6 times higher than OR for even the strongest clinical predictor (invisible/unresected LGD).

Low-grade dysplasia is notoriously hard to risk-stratify, even with some known risk factors for later cancer (e.g., size, multifocality). We hypothesized that the true culprit was underlying altered genomes of cells and conducted a large case-control study to test this at index LGD timepoint.

Low-coverage whole genome sequencing of low-grade dysplasia strongly predicts advanced neoplasia risk in ulcerative colitis Background The risk of developing advanced neoplasia (AN; colorectal cancer and/or high-grade dysplasia) in ulcerative colitis (UC) patients with a low-grade dysplasia (LGD) lesion is variable and dif...

Today our study using copy number alterations (CNAs) from low-coverage WGS to predict FUTURE colorectal cancer in patients with IBD is published, details in thread! 1/5
gut.bmj.com/content/earl...

Incomplete human reference genomes can drive false sex biases and expose patient-identifying information in metagenomic data - Nature Communications Here, the authors apply novel pangenome indexes created from diverse human references to facilitate computationally efficient removal of host genetic information, preserving biological accuracy and pr...

Our work on filtering host DNA from WGS data is also out today in @naturecomms.bsky.social! Congrats to first authors Caitlin Guccione and Lucas Patel - their methods were used in our new preprint and are freely available to use on your own data 🧬
www.nature.com/articles/s41...

This is just the start of our work to tease apart relationships between host-microbe clonal populations in BE progression from an evolutionary perspective. Thanks for the fun collaboration with labs @ucsdmedschool.bsky.social @ucsandiego.bsky.social @fredhutch.bsky.social !

Finally in model simulations, we found that assuming widely varying growth rates across taxa best reflected EAC data, implying that selection pressures may influence largely niche-based population dynamics in the tumor microenvironment.

We found evidence of mostly neutral processes in normal and precancerous tissue, with the exception of Helicobacter pylori which deviated from a neutral expectation in BE patients who did not progress to EAC. For the EAC microbiome, patterns seem largely non-neutral.

Starting with Caitlin’s pipeline to deplete WGS data of host DNA (www.researchsquare.com/article/rs-4721159/v1 in press Nat Comms), we modeled microbial population dynamics during normal esophagus --> GERD --> BE --> EAC progression

How do microbial communities assemble during precancer and cancer evolution? Excited to share our new preprint on modeling of microbial evolution in Barrett's esophagus and esophageal adenocarcinoma! 🦠 Led by PhD student Caitlin Guccione #barretts #mathonco more in 🧵
www.biorxiv.org/content/10.1...