- Over 50 yrs since the discovery of protein kinases, 80% of human kinases still have ≤20 known substrates, and many are “dark.” I'm EXCITED to announce our new work towards solving this- combining (1) deep learning with (2) proximity proteomics in vivo! ➡️ www.biorxiv.org/content/10.1...
- (1) Our deep learning model, KolossuS, leverages large protein language models to learn the “grammar” of kinase–substrate recognition, achieving SOTA accuracy and performance across the mammalian kinomes. It is well-calibrated across kinase families, enabling real-world usage.Apr 28, 2025 22:00
- (2) It is insufficient to simply predict kinase specificities in vitro; kinases must also meet their substrates in cellular contexts. We tag endogenous kinases with TurboID in mouse brain, capturing substrates within ~10 nm in living animals.
- This allows for the capture of nearby phosphorylated substrates for kinases in vivo. KolossuS accurately interprets this data. We validate our framework by capturing state-dependent substrates, focusing on Sik3 and its differential substrates in response to sleep deprivation in vivo. 😳 vs 😴
- We find known and novel Sik3 substrates that shed new light on the molecular mechanisms of sleep by combining KolossuS and CRISPR-based BioID.
- The BEST part: This would not have been possible without the close (and SO FUN!) collaboration of my lab with @rohitsingh8080.bsky.social and the lab of Masashi Yanagisawa.
