preciseTAD, a transfer learning framework for precise 3D domain boundary prediction, is finally out! It learns boundary-epigenomic associations at the Hi-C resolution and predicts the probability of each base being a boundary doi.org/10.1093/bioinf…

In our latest episode of Roman Cheplyaka | the bioinformatics chat, we talk with Žiga Avsec about research in academia vs industry, Enformer, and deep learning libraries! Great to hear about the work directly from the source. Hope other people enjoy our conversation! bioinformatics.chat/enformer

Our paper is online! We (as in Nader) assayed the human proteome for transcriptional activators with a pooled screen and found over 200 activators. These included both well-characterized activators and unknown proteins. sciencedirect.com/science/articl… 1/6

#ScienceForUkraine If you are a postdoc, PhD or Master's student interested in Cancer Genomics and would like to join our lab (bbglab) for an internship contact me. List of labs supporting Ukrainian Scientists. docs.google.com/spreadsheets/d… 🇺🇦#ScienceForUkraine #ScienceForUkraine 🇺🇦

This is an excellent paper for multiple reasons. Firstly, very important problem and a very significant contribution scientifically. But I'm going to focus on another aspect. IMHO, it's an excellent case study showcasing best practices for papers using ML for genomics. 1/

One huge advantage of deep learning (vs classical ML models) that is not often discussed is *modularity*: One can download pre-trained models, glue them like Legos and fine tune them end-to-end because gradients flow through. (1/n)

We've known for a long time that a large fraction of noncoding GWAS hits are not currently explained by eQTLs. Why?? I'm excited to share our new preprint on this! Take home: GWAS and eQTL assays maximize power for different types of variants. (1/n) biorxiv.org/content/10.110…

VGG, U-Net, TCN, ... CNNs are powerful but must be tailored to specific problems, data-types, -lenghts & -resolutions. Can we design a single CNN that works well on all these settings?🤔Yes! Meet the 𝐂𝐂𝐍𝐍, a single CNN that achieves SOTA on several datasets, e.g., LRA!🔥

‼️ ONLINE NOW Nature Genetics 📰 A sequence-based global map of regulatory activity for deciphering human genetics 🧑🏿‍🤝‍🧑🏻 Jian Zhou Olga Troyanskaya and team 👇🏻 go.nature.com/3ACPk3m

Using a sequence-based deep neural network, scBasset facilitates various tasks of single-cell ATAC-seq analysis in a unified framework. Han Yuan David Kelley Calico nature.com/articles/s4159…

Nested epistasis enhancer networks for robust genome regulation science.org/doi/10.1126/sc… This research between our lab, Keji Zhao (NIH), Wing Wong (Stanford), reveals an epistasis enhancer network for gene expression robustness, and predicting non-coding genome in disease risk.

Impactful words by Carolyn Bertozzi today on how diversity in her trainees enabled creativity, not playing by the rules, and ultimately the work that built the field of bioorthogonal chemistry. Diverse science leads to better science!

How do the #genome, #epigenome and #transcriptome co-evolve in colorectal #cancer? In the second of two papers in @nature, we find gene expr is not strongly heritable & evaluate "driverness" of muts directly in patient tumours nature.com/articles/s4158… Andrea Sottoriva #EPICC2 1/7

Full house during the third joint meeting organized by Oncode and @TUeindhoven: Sharing knowledge on AI & Big Data. We are discussing the statistical and mathematical analysis of large biological/clinical data sets and its potential use for cancer research.

The fundamental importance of basic science nature.com/articles/d4158… This week's nature editorial

People think Machine learning can solve all the biology problems. In reality, see the MEME. That said, ML is very powerful when used correctly. 8 papers to avoid ML pitfalls in biology👇 🧵

Our lab is working on (single-cell) eQTL mapping in various tissues. We have a PhD position available that concentrates on studying how genetic variation affects brain gene expression. Interested? Please get in touch with us!

🆕nature Deep learning #AI during surgery, using low cost, intraoperative nanopore sequence, to rapidly classify tumors in <90 minutes nature.com/articles/s4158… @jeroen_deridder

The first preprint of our PERICODE consortium: MPRA-trained deep learning provides insight into the regulatory logic of promoters and transcription factors. biorxiv.org/cgi/content/sh…

Our review is finally out! 🙂🙂