Markus List

A flowery surprise at our @tum.de campus Freising yesterday. Congratulations to all students who celebrated their graduation.

Yes indeed, are you here as well :-)

En route to visit the @cosybio-uhh.bsky.social lab in Hamburg who are kindly organizing the latest @repo4eu.bsky.social WP2 workshop. Looking forward to discussing the refinement of our computational pipelines for drug repurposing. Hope the train will not be too much delayed...

🧬🖥️Drug response prediction is a machine learning challenge with immense potential for precision medicine. Our latest preprint introduces DrEval, a comprehensive benchmarking framework to evaluate state-of-the-art methods, uncover widespread issues, and guide the development of more robust models.

Had a great time in Innsbruck. The scenery here with the mountains in the background is always impressive, even when the weather is not so nice. Thanks @francescafinotello.bsky.social for inviting me!

webinar description

For those of you who are not in Innsbruck to see me today, you might instead listen to @judith-bernett.bsky.social at the @iscb.bsky.social NetBio webinar!

🔗 Attend at ISCB Nucleus: iscb.junolive.co

📍 If you’re not an ISCB member, register for access to ISCB Nucleus: lnkd.in/gMhrKGJz

Guiding questions to avoid data leakage in biological machine learning applications - Nature Methods This Perspective discusses the issue of data leakage in machine learning based models and presents seven questions designed to identify and avoid the problems resulting from data leakage.

I believe the seminar is offline only. My focus is not data privacy (also very important!), but on inflated performance estimates due to methods learning illegitimate shortcuts. If you'd like to know more, we have written a perspective article with guiding questions: www.nature.com/articles/s41...

Colorful liquid flowing from one bottle into another, as an illustration for (data) leakage.

Traveling to Innsbruck by invitation of @francescafinotello.bsky.social to talk about data leakage, a widespread issue in biomedical machine learning applications. I'll talk about challenges in protein-protein interaction (doi.org/10.1093/bib/...) and drug response prediction (upcoming preprint!).

Benchmarking algorithms for spatially variable gene identification in spatial transcriptomics AbstractMotivation. The rapid development of spatial transcriptomics has underscored the importance of identifying spatially variable genes. As a fundament

Benchmarking algorithms for spatially variable gene identification in spatial transcriptomics 🧬🖥️ academic.oup.com/bioinformati...

Weihenstephan Bioinformatics Symposium 2025: More than 75 scientists from Bavaria and the world came together to share talks and create new synergies. It was great to host this event, for those who missed it: The next edition is planned for 2027 😉

I am so happy to see this manuscript finally out!!! We review and discuss all analysis steps in long reads transcriptomics. Hope the community finds this useful! Hugo thanks to @carolinamonzo.bsky.social and @tianyuanliu.bsky.social for the huge work!!! @longtrec.bsky.social @hitseq.bsky.social

Greetings from Palermo! @en-coding.bsky.social, @a-dietrich.bsky.social, @itisalist.bsky.social, Serafina Reif, Nico Trummer & Kamila Kwiecien are united here at the occasion of the MyeInfoBank COSTAction: Converting Molecular Profiles of Myeloid Cells into Biomarkers for Inflammation and Cancer

It was a pleasure having you, Ryu! Thanks so much for your talk and visit.

📍Welcome to our presentation round of the DaiSyBio members! Every week, you will get to know someone from our lab.
The start is done by @itisalist.bsky.social who heads the group. Markus joined TUM in 2018 and became a W2 tenure track associate professor in 2023. More members are about to follow! 📍

DaiSyBio | LinkedIn DaiSyBio | 323 followers on LinkedIn. DaiSyBio is a research group created in December 2023 as part of the TUM School of Life Sciences. Cutting-edge expertise in biology, medicine and computer scienc...

By the way, we also have a lab account on bluesky now. Follow @daisybio.de to get more news about our activities. We are also on LinkedIn: www.linkedin.com/company/dais...

Johannes Kersting Awarded an ASAPBio Poster Prize for RExPO Contribution The ASAPBio poster competition is an initiative that champions open science by highlighting innovative research and the early sharing of results. This year’s competition emphasized transparency, colla...

Congrats to our PhD student Johannes Kersting for winning the ASAPbio poster prize for his contribution to the @repo4eu.bsky.social #REXPO24: blog.scienceopen.com/2025/01/joha...

We followed up on our previous work, where we showed that predicting protein-protein interactions from sequence alone yields random performance when data leakage is accounted for. In this new preprint, we show that ESM2 embeddings raise the bar to 0.65 accuracy independent of the model architecture.

Hey young investigators, use this opportunity to learn about the COST action MyeInfoBank and listen to a nice talk by our fabulous @a-dietrich.bsky.social

The whole DaiSyBio team wishes you happy holidays - more to come in 2025 ✨ (Nota bene: This is not an AI generated picture)

CoBiNet - Context-specific, bias-reduced protein-protein interaction networks Methods for context-specific and bias-reduced protein-protein interaction networks.

Another issue is that static PPI networks do not reflect dynamic changes. Continuing our fruitful collaboration with @dbblumenthal.bsky.social & @michaelschaefer.bsky.social in the project www.cobinet.ai, generously funded by @klaus-tschira-stiftung.de, we explore strategies to improve PPI networks.

We realize here that PPI networks might look different than we think. Normalizing for study bias promotes proteins such as chaperones and demotes disease-relevant proteins, showing that disease-attention bias could potentially be mitigated.

In this paper, we challenge the widely held view that the degree of distribution of protein-protein interaction networks follows a power law. Indeed, we could show through simulation that this property can be explained by a combination of study attention bias and false positive detections alone.

Emergence of power-law distributions in protein-protein interaction networks through study bias

Incredibly happy to finally see our manuscript "Emergence of power-law distributions in protein-protein interaction networks through study bias" published in @elife.bsky.social. doi.org/10.7554/eLif... It's been a long but fun journey with @dbblumenthal.bsky.social and @martinschaefer.bsky.social

NiCo is out! Our new algorithm for spatial transcriptomics data analysis predicts the crosstalk of cell types co-localized in tissue niches and sheds light on signalling mediators and downstream effects of cell-cell interactions by inferring covarying gene programs.

www.nature.com/articles/s41...

Looks like yet another case of overhyped results due to poor #bioMLeval evaluation of deep learning models -this time deep docking methods - specifically DiffDock. Look forward to the DiffDock authors response. But dont see any major flaws in this critique. Conclusion is REALLY worth reading!

Method of the Year 2024: spatial proteomics - Nature Methods Approaches for profiling the spatial proteome in tissues are the basis of atlas-scale projects that are delivering on their promise for understanding biological complexity in health and disease.

I could not be more thrilled to announce the Nature Methods @naturemethods.bsky.social Method of the Year is Spatial Proteomics! Please see our editorial as a roadmap to the fantastic content in this special issue! www.nature.com/articles/s41...

Overview figure of the seven guiding questions addressed in manuscript regarding (1) Study bias, (2) Label distribution, (3) Data shift, (4) Inter-sample similarities, (5) Unavailable features, (6) Features as illegitimate surrogates, (7) Pretraining

🧬🖥️Tranferring my papers, Pt. 2: "Guiding questions to avoid data leakage in biological machine learning applications" doi.org/10.1038/s41592-024-02362-y, which I published with @itisalist.bsky.social @dbblumenthal.bsky.social @romanjoeres.bsky.social, D. Grimm, F. Haselbeck, and O.V. Kalinina. 🧵1/20

Overview figure summarizing the major analyses of the paper. In the setting employed by published methods, performances are strongly inflated regardless of model and dataset. When the positive training dataset is randomly rewired, there is only a slight decrease in performance. When the models are evaluated on unseen proteins that are unsimilar to the training proteins, performances become random.

🧬💻Transferring my paper posts here, starting off with "Cracking the black box of deep sequence-based protein-protein interaction prediction" doi.org/10.1093/bib/..., which I published together with @itisalist.bsky.social and @dbblumenthal.bsky.social. So what was it about? 1/13 🧵

ChIP-DIP maps binding of hundreds of proteins to DNA simultaneously and identifies diverse gene regulatory elements - Nature Genetics ChIP-DIP (ChIP done in parallel) is a highly multiplex assay for protein–DNA binding, scalable to hundreds of proteins including modified histones, chromatin regulators and transcription factors, offe...

Gene regulation involves thousands of proteins that bind DNA, yet comprehensively mapping these is challenging. Our paper in Nature Genetics describes ChIP-DIP, a method for genome-wide mapping of hundreds of DNA-protein interactions in a single experiment.
www.nature.com/articles/s41...