Twitxy Saved

Gunnar Rätsch

@gxr

I love large biomedical data.
@gxxxr.bsky.social
Zurich, Switzerland ratschlab.org Joined April 2009
416 Following 3K Followers 1K Posts

Latest posts 5

Gunnar Rätsch @gxr
Thanks for organizing! Such challenges are very important to have in order to distill whats "good sales" and whats good "performance". However, most performance measures have limitations & the real test is to use the method for scientific discovery or translational applications.
Our Autoimmune Disease ML Challenge drew in ~1,000 participants from 62 countries to solve the problem of predicting gene expression from pathology images. We’re thrilled to present the top performers of Crunches 1 and 2 – learn more:
ericandwendyschmidtcenter.org/updates/machin… @broadinstitute
View quoted post on X
Gunnar Rätsch @gxr
Congrats Kalin @nonchevk! Great work!
Top Performers:

🏆 @nonchevk – 1st Crunch 1, 2nd Crunch 2 (@ETH_en)
🏆 Manfred Seiwald – 2nd Crunch 1 (University of Salzburg)
🏆 Team PathBio – 3rd Crunch 1 – Sen Yang (@Stanford), Jinxi Xiang (Stanford), Wei Yuan (@SCUCN), Yijiang Chen (Stanford), Xiyue Wang (Stanford)
View quoted post on X
Gunnar Rätsch @gxr
This manuscript on the reproducibility of splicing neoepitope prediction took us a while. We find that subtle choices in how one exactly generates and filters neoepitope candidates, creates major differences in the number and identity of the neoepitopes. biorxiv.org/content/10.110…
www.biorxiv.org Splicing neoepitope prediction is sensitive to methodological differences Motivation Cancer-specific neoepitopes may arise from abnormal splicing in the transcriptomic landscape (alternative splicing neoepitopes, ASNs), leading to divergent proteins with high potential...