🎙️ Next up Dec 2 in VESS!

Thea Schulze (Lindorff-Larsen Lab): Predicting mutated protein abundance @tkschulze.bsky.social

Taylor Mighell (Lehner Lab): Massive mutagenesis to understand GPCRs @taylor-mighell.bsky.social

🔗 More info at varianteffect.org/seminar-series
@varianteffect.bsky.social

Coming up next in the VESS (Nov 4):

Population genetics × variant effects:

🧬 Nikhil Milind (Stanford) on gene dosage and complex traits @nikhilmilind.dev
🧬 Leslie Smith (U Florida) on equitable ML in cancer genomics

www.varianteffect.org/seminar-seri...
@varianteffect.bsky.social

VESS is happening tomorrow (Oct 7). See you there!
First speaker: Shelby Hemker (Dr. Jacob Kitzman Lab, University of Michigan)
Second speaker: Karl Romanowicz (Dr. Calin Plesa Lab, University of Oregon) @kroman.bsky.social
Link: www.varianteffect.org/seminar-seri...

We applied Cosmos to 3 DMS datasets:
Kir2.1: abundance → surface expression
PDZ3: abundance → CRIPT binding
KRAS: abundance → RAF1_RBD binding
Cosmos clearly separates direct binding residues from those with indirect effects.
[6/n]

How does it work?
Cosmos aggregates mutation effects by position, learns interpretable causal graphs via Bayesian model selection, and outputs residue-level direct vs indirect effects.
[5/n]

Cosmos is a Bayesian framework that performs residue-level causal inference to decouple how mutations influence upstream vs downstream functions.
No need for detailed biophysical models—just stats and data.
[3/n]

Multi-phenotype DMS experiments are revealing how mutations impact different protein functions.
But these phenotypes are often causally linked—e.g., when measuring activity, we may also capture effects propagated from abundance.
So how do we tell what’s direct vs indirect?
[2/n]

Perfect first day: receiving a set of new pipettes! Can’t wait to do more cool experiments with the lab for the next few years. @willowcoyote.bsky.social

Thrilled to share that I just successfully defended my PhD! Thanks to my committee, collaborators, and everyone who’d supported me throughout my seven years at UCLA. A special thank you to my PI Harold for his incredible mentorship! @hjp.bsky.social

Bye SF. Guess it's time to stop pretending I’m a biologist.

Why is it so hard to include an actual data file in the paper? Attaching one in a pdf is not the “nature” way.

Real data shows Rosace is very robust. It catches many experimentally validated functionally deviant mutations while flagging few baseline mutations as deviant. (high power and low false discovery, in stats-speak.) (6/n)

Rosace is the first DMS model to incorporate position. It implements a hierarchical model that parameterizes each variant's effect as a function of the position effect, providing a way to incorporate both position-specific information and shrinkage into the model. (5/n)

Intuitively, missense mutations at the same amino acid location would likely have similar effects (e.g. missense mutations at a critical location breaks the protein), but existing tools have not considered this “positional effect”. And Rosace comes to rescue! (4/n)