Florian Wünnemann

Check out the first release of our @nf_core @nextflowio pipeline for processing @ResolveBioSci Molecular Cartography data🤩! Had a lot of fun working on this with @KresBestak 🎉. Reach out to us if you have any questions or need help getting started!

Aggregating Residue-Level Protein Language Model Embeddings with Optimal Transport 1. This study introduces a novel protein representation method using sliced-Wasserstein embedding (SWE) to aggregate residue-level outputs from protein language models (PLMs) into fixed-length, protein-level embeddings. SWE addresses limitations of traditional average pooling by capturing essential sequence information often lost in standard aggregation methods. 2. SWE interprets token-level embeddings as samples from a distribution and maps them onto a reference set using optimal transport, creating embeddings that retain critical information from variable-length protein sequences. This approach significantly outperforms average pooling, especially for longer sequences, by preserving details vital for accurate predictions. 3. Using SWE, small PLMs achieved comparable or superior performance to larger models with average pooling across several tasks, including drug-target interaction (DTI), enzyme commission (EC) prediction, and subcellular localization, making PLM use more accessible for researchers with limited computational resources. 4. This approach enhances interpretability, as the learned reference embeddings represent key features of protein-level tasks, providing insights into protein structure-function relationships and offering a scalable solution for protein-level prediction tasks. @rohitsingh8080 💻Code: github.com/navid-naderi/P… 📜Paper: biorxiv.org/content/10.110… #ProteinLanguageModels #MachineLearning #Bioinformatics #OptimalTransport #ProteinRepresentation

@design_proteins Amazing, thanks! I was wondering which steps in the pipeline would affect ranking. For example, do novel binders have lower folding confidence due to no/bad MSA? Are the metrics we use inherently biasing for existing sequences?

@flowuenne Just checked now, and I see 217/400 had hits to the PDB, leaving 183 de novo designs

Design selections from @adaptyvbio EGFR binder competition index 0-99 are the top ranked by score, rest are handpicked by adaptyv interesting to see the trend towards smaller designs in the top 100 I blasted all designs against the PDB and ...

@design_proteins Thanks for the insight from your follow up analysis @design_proteins! Did you also check how many across all 400 designs were novel? I was also thinking that the selection metrics was rather highly biasing for modification of known binders vs novel designs...

my best was # 114. I tried both BindCraft and RSO to design binders and eventually found RSO to give both my best and fastest designs. Once I found good hits, I optimized them with RSO by inputting the sequence back into the design process. I'm excited to see the results!

@_judewells @adaptyvbio Awesome to see so many different approaches to this competition by different people! How did you run your experiments? Did you use any specific tooling or did you have access to a uni HPC with "free"ish gpu access?

Just submitted some designed proteins to @adaptyvbio EGFR binder competition. My strategy was Chai1 optimization on a diverse set of known EGFR binders shown below. Lower 2 are antibody FABs (necitumumab and panitumumab) which are too large for the competition requirements so...

@GrapotteM @tallphil It was great to see you and hang out @GrapotteM! Your talk was honestly one of the highlights of the summit for me 👍🚀! Hope to see you at thr next summit or maybe earlier in Montreal or Barcelona ☺️!

@tallphil Thank you for being such an important part of the community. The hackathon & summit had such great vibes and it reflects the spirit of the people who made this possible, it feels a bit surreal to go back to the lab on Monday!

Very tired but very happy after a fantastic week. Thank you to everyone who came and joined us at the #NextflowSummit, as well as to everyone who made it possible - you know who you are!

@design_proteins @modal_labs @chrisfrank662 Looking forward to see how these RSO designs fare in the leader board! Did you submit any yet?

@flowuenne @modal_labs @chrisfrank662 Thanks! I got 3 passing designs in ~2 minutes on an H100, and H100 pricing is $0.001267/second on modal. So 3designs*(1/120seconds)*(1second/$0.001267)*$500= ~9865 designs! I do plan to check design time vs cost to find the optimal GPU though

Finally spun up my first @modal_labs app with the impressive RSO binder design method from @chrisfrank662! It's bare bones but works, try it out github.com/coreyhowe999/R… I produced this EGFR binder in ~2 mins on an H100, we'll see how fast I chew through the $500 credits...

Happy to release a concise review of recent progress in sequence-structure co-generation for protein design. with our amazing intern @WChentong + @SarahAlamdari Carles Domingo-Enrich and @avapamini

Writing the Response to Reviewers document

Really excited for this talk by @GrapotteM about how to use Nextflow for training deep learning models! Really great to see Nextflow adoption in many new areas & fields!

It’s time for the #NextflowSummit BCN speaker series! 🎉 Let's welcome our first speaker Ziad Al Bkhetan, Product Manager at @AusBiocommons 🎤 He’ll present "The Australian ProteinFold Service: Interactive Prediction & Visualisation of Protein Structure" summit.nextflow.io/2024/barcelona…

Happy to share Novae, our new foundation model for spatial transcriptomics data 💫 Load a pre-trained model and start using Novae in two lines of code! biorxiv.org/content/10.110… github.com/MICS-Lab/novae

Are you looking for job opportunities where your #Nextflow skills are valued? Check the list we've compiled below. #Jobs #Opportunities

@strnr Looks like an interesting pipeline, hope any new approaches can be implemented in @nextflowio nf-core/spatialvi. I do love the name, well done 👏!

SpatialOne: end-to-end analysis of visium data at scale academic.oup.com/bioinformatics… 🧬🖥️ github.com/Sanofi-Public/…

#scverse2024 conference is on! We, @scverse_team, couldn’t be more excited to meet everyone here! We all know each other and finally get to hang out together. Amazing moments for the “young” science!

Come join us at 5Prime as we build software and algorithms to bring more medicines to patients. Senior Full Stack Engineer posting: linkedin.com/jobs/view/4021…

I remember using @nextflowio for the first time and experiencing the portability, reproducibility and parallelization it provided. With Seqera you can now take this one step further and optimize your workflow resources, manage teams and scale your workflow executions! 🚀

Still time until 15th September to submit an abstract to join us in Berlin in the @SpatialBio conference 👇

Watch the talk from Joseph Rich, grad student in @lpachter's lab, from BioC2024 timestamp ~12:30 youtu.be/T6cOyKHlxpE?si…

Processing raw data using efficient, scalable @nextflowio pipelines and then performing interactive data analysis is a common use case for scientists! @SeqeraLabs Data studios brings the tools you love next to your data in the cloud 🤩☁️.