Jeff Saucerman

More Riprap #hiking

Rain fog and leaves turning along Riprap. One of my favorite trails in Shenandoah NP. #hiking

These show that neuregulin-1 induces cardiomyocyte Feret elongation via sustained PI3K signaling, while transient p38 signaling increases cell size but not shape. This study leverages experimental-computational methods to dissect how signaling dynamics distinctly regulate cell size vs. shape. 3/3

We first characterized the proteomic, gene expression, and phenotypic responses of cardiomyocytes (including our new cell shape metric Feret elongation) to diverse ligands. We integrated this data with data-driven and mechanistic computational models, validated with new perturbation experiments. 2/

How do signaling pathways differentially control cell size vs. shape? In a new preprint led by Alice Luanpaisanon, we iterate between experimental and computational methods to discover such pathways. 1/ www.biorxiv.org/content/10.1...

I would say that tracks but...

More pics

Spooky fall day in Shenandoah NP

Finally, check out the eye candy! Awesome multi-channel live-cell videos analyzed with machine learning segmentation and tracking. Shows the complex fates that cells commit to. 3/3

Inhibition of caspase-3 changes a cell's bias from apoptosis to hypertrophy. Initial cell and nuclear morphological features also bias a cell's decisions. These results demonstrate that the single-cell dynamics of cardiomyocyte growth and death are highly heterogeneous and dynamic. 2/3

Paradoxically, heart failure involves both cell death and growth. How does a cell decide? In a paper at JMCC Plus led by Bryan Chun and Lavie Ngo, we find that rather than dying following hypertrophy, stressed cardiomyocytes commit early to either grow or die.
www.sciencedirect.com/science/arti...

Thanks! The Point, Rose River Loop, Big Meadows

More hiking pics

Nearing sunset, Shenandoah NP #cville

@j-muncie-vasic.bsky.social and @benoitbruneau.bsky.social show that MEF2C is not a one trick pony- it distinctly regulates different regions of heart tube formation. Glad @aclark.bsky.social and I could contribute gene regulatory networks that predict mechanisms.

Nice to walk out of my backyard onto the #Crozet Connector Trail. #cville

Further, pharmacologic and postnatal genetic inhibition of DYRK1A enhanced cardiomyocyte cycling and cardiac functional recovery after MI. Overall, our findings reveal network mechanisms by which a small molecule inhibitor drives cardiomyocyte cycling and post-MI functional recovery. 3/3

We used network modeling to predict mechanisms by which DYRK1A inhibition induces cardiomyocyte cell cycling. We validated these predictions using a recently described small molecular inhibitor of DYRK1A by imaging and RNA sequencing of cultured cardiomyocytes. 2/3

Happy to share our latest preprint, "Network Modeling Predicts How DYRK1A Inhibition Promotes Cardiomyocyte Cycling after Ischemic/Reperfusion Injury", in collaboration with Matthew Wolf's lab, led by Bryce Murillo and Alex Young! www.biorxiv.org/content/10.1...

Cooling off at Sugar Hollow #cville

More pics

Sunset showers at Shenandoah NP

Thank you!

Benchmarking of signaling networks generated by large language models Computational models of signaling networks provide frameworks for predicting how molecular cues guide cell decisions. But they are typically limited by manual curation from incomplete literature. Here...

Our pipeline provides a proof-of-concept and benchmarking framework for LLM-generated models of signaling networks. But their accuracies do not yet approach that of literature-curated networks. We appreciate your feedback on the preprint! Jeevan and Ben 👏. 3/3 www.biorxiv.org/content/10.1...

We find that LLM-generated networks reconstruct 24-58% of the structure of literature-curated networks. When challenged to predict perturbation responses from independent literature, LLM-generated networks validate at 5-26% compared to 77-95% validation accuracy with literature-curated models. 2/

Do large language models understand how cells work? To find out, we built a pipeline that generates logic-based models of signaling networks using general purpose LLMs GPT, Gemini, or Claude. We benchmarked LLM-generated networks against 3 large-scale literature-curated and validated networks. 1/

I love that beach! Have a handful of pebbles on my desk

And Baltimore science museum!

Great interactions, crabcakes, and science at #BCVS2025 in Baltimore