Bin Zhang

Congratulations to Professor Bin Zhang, who has been honored as “Committed to Caring”. The C2C program is a student-driven initiative that recognizes outstanding professors who extend this dedication beyond the classroom. chemistry.mit.edu/chemistry-news…

Pretraining implicit solvent, or coarse grained, models is difficult, limiting the transferability of the resulting ML force fields. However, Justin just figured out a novel method for pretraining using protein language models. More can be found at: arxiv.org/abs/2601.05388

Our collaborative paper with @XuhuiHuangChem is now published in Biophy J (authors.elsevier.com/a/1mEr91SPTBoOU). I learned a lot about kinetic modeling via this collaboration, and it can be quite illuminating for understanding chromatin folding!

Thrilled to share our new JCP editorial (shorturl.at/6mKdY), co-authored with @tamar_schlick, introducing the special collection “Chromatin Structure and Dynamics: Recent Advancements.” Huge thanks to all the outstanding contributors!

Our paper is now online at Sci Adv: science.org/doi/10.1126/sc… See also the accompanying news: news.mit.edu/2025/with-gene…

Hmm, I assume these posts were made by AI? Anyway, very impressive summary of our latest preprint.

Check out our recent manuscript that studies the interactions between small molecules and biomolecular condensates with all-atom simulations: pubs.acs.org/doi/abs/10.102…

Our findings shed light on the significant regulatory roles of bio condensate and DNA linker length and help bridge the gap between in vivo and in vitro observations.

Nucleosome condensate and linker DNA alter chromatin folding pathways and rates [Qiu et al, 2024] biorxiv.org/content/10.110… ▶️residue-level coarse-grained models; non-Markovian dynamics ▶️10n bp DNA linker lengths favor zigzag fibrils ▶️10n+5 bp chromatin loses unique conformations

Scaling Graph Neural Networks to Large Proteins 1. This paper introduces the DISPEF dataset, specifically designed for benchmarking graph neural networks (GNNs) on large, biologically relevant proteins. DISPEF contains over 200,000 proteins with implicit solvation free energies, a key challenge for molecular modeling. 2. A major innovation is the introduction of a multiscale architecture called “Schake,” which enhances GNN performance on large proteins by incorporating both short-range and long-range interactions, ensuring computational efficiency and transferability across protein sizes. 3. The Schake model leverages two types of message-passing layers: a more accurate SAKE layer for short-range atomic environments and a more efficient SchNet layer for long-range alpha carbon interactions. This mixed design significantly improves both accuracy and speed. 4. DISPEF provides a diverse chemical environment, including both folded and disordered protein regions, making it an ideal dataset for testing GNNs. The inclusion of many-body solvation free energies as targets pushes the limits of model accuracy and generalization. 5. Benchmark results show that Schake consistently outperforms existing GNNs on energy and force predictions for large proteins, while maintaining superior computational efficiency, reducing inference times by up to 2.7× compared to prior models. 6. A key insight from the study is that increasing the cutoff distance in GNNs improves model transferability to larger proteins, but this needs to be balanced against computational cost, which Schake addresses with its hybrid architecture. 7. This work highlights the importance of datasets like DISPEF for driving future GNN innovations and optimizing models for real-world applications like protein folding and drug discovery, where large proteins are common. 8. The paper provides valuable insights for advancing GNN architectures in computational biology, emphasizing the need for both accuracy and efficiency in handling the vast complexity of biomolecular systems. @binzmit 💻Code: github.com/ZhangGroup-MIT… 📜Paper: arxiv.org/abs/2410.03921

@Ella_Maru Great question! ChromoGen was trained with only GM12878 data, and we showed that the prediction results for IMR90 cells were equally accurate.

@binzmit Congratulations Dr. Zhang! Have you tested ChromoGen's prediction accuracy across various cell types and tissues beyond the training set?

I am excited to share our latest #preprint: ChromoGen: Diffusion model predicts single-cell chromatin conformations researchsquare.com/article/rs-463… As the title suggests, we achieved de novo prediction of 3D chromatin structures using DNA sequence and ATAC-seq using AI.

Our explicit ion paper is now online at Elife: elifesciences.org/articles/90073. But who has time to read anyway? Luckily, @XingchengLin made an excellent video that you can watch at: youtube.com/watch?v=4l6G0y…

Our review on chromatin organization is online at the Annual Review of Biophysics! Spoiler alert, we took a more physical chemist's perspective on the problem. annualreviews.org/doi/pdf/10.114…

Thanks to the efforts of our group members. This new work provides a new role of micropolarity in condensates.

You can also listen to an audio summary here: sciencecast.org/casts/0msn7rgv…

Interested in IDPs and biomolecular condensates? Do you ever wonder why IDPs adopt low-complexity domains? How do you map a given protein sequence into stickers and spacers? We attempt to address these questions in our latest preprint. biorxiv.org/content/10.110…

A fruitful collaboration with Bin Zhang @binzmit reveals the nanometer scale interfacial environment of phase separated condensates. In @eLife: Frustrated Microphase Separation Produces Interfacial Environment within Biological Condensates doi.org/10.7554/eLife.…

We welcome new postdoc, Joe Paggi, and grad students Camryn Carter and Ivan Riveros to the group. I am very much looking forward to working with you all.