BioMoDes Newsletter -- Issue #01

Here are the latest tools (from last week) for biomolecular modeling and design.

1. Intro

Welcome to the very first issue of the BioMoDes Newsletter and Top Reads, a newsletter where I share with you the latest tools and my top-reads in Biomolecular Modeling and Design, focusing on computational structural biology and drug design.

Since my research works have always involved the use of SOTA tools and approaches in computational structural biology, I tend to follow the literature very closely. That means I frequently come across cool tools and methods.

I try to systematically keep track of and organize what I read/come across to make it easier for me to go back to the materials when needed, especially with the recent surge in the number of tools and methods for biomolecular modeling and design. Then I thought it might be helpful to share these resources with other people who might potentially find them useful—reason for the BioMoDes repository and newsletter.

Ok.

A lot of new tools came out since I announced BioMoDes, so I will only be sharing some of the most important ones that I noted and will be carrying others over to the next issue. I will also share some of my top reads for the past week.

2. New tools over the last couple of weeks

0. FoldFlow-2: A flow matching model for protein structure generation that outperforms RFdiffusion on multiple metrics and tasks. As the name suggests, FoldFlow-2 is an improved and newer version of FoldFlow, both developed by DreamFold, a spin out from the lab of Yoshua Bengio. FoldFlow-2 leveraged PDB structures and high-quality artificial structures for training. Read more...

1. RNADiffFold: A generative model for RNA secondary structure prediction. RNADiffFold leverages neural networks from RNA-FM and UFold models for feature extraction. Read more...

2. ProTrek: A tri-modal protein language model that simultaneously models protein sequence, structure, and function (SSF). ProTrek is a protein search tool that uses a natural language search interface and offers 9 pairwise SSF searching tasks (all 9 pairwise combinations of the SSF). In protein-protein search tasks, ProTrek is 100x faster than Foldseek and MMseqs2, and surpasses ESM-2 in 9 out of 11 prediction tasks. Read more...

3. Proteus: A diffusion model for de novo protein backbone generation that surpasses RFdiffusion on multiple metrics. Proteus outperforms other leading structure generation models (RFdiffusion, Genie, FrameDiff, and Chroma) on designability, efficiency (speed), and diversity of designs. On short sequences (100-300 residues), Proteus matches RFdiffusion in scRMSD; but on longer sequences, Proteus performs better. Proteus also showed excellent wet experimental results. Read more...

4. RNAFlow: A flow-matching model for simultaneous design of RNA structure and sequence design conditioned on protein structure and sequence. The denoising network of RNAFlow comprises: 1. an RNA inverse folding model, and 2. a pre-trained RosettaFold2NA network. Read more...

5. AlphaKnot 2.0: A database and webserver for measuring entanglements in AlphaFold models. AlphaKnot helps to determine the knot type in predicted and experimental models. Basically, AlphaKnot 2.0 offers two functions: 1. analysis of knotting in predicted or user-provided models, and 2. a repository of knotting in AlphaFold models for >200 million models. Read more...

6. Genie 2: A generative model for structure-based protein design that sets a new state-of-the-art by outperforming existing models on various metrics. Genie 2, a model by AlQuraishi's team, builds on Genie (version 1) to expand the structure space (size and diversity) captured by the model. It surpasses existing methods in diversity, designability, and novelty. Motif scaffolding task has also been added to the capabilities of Genie 2, outperforming RFdiffusion. Read more...

3. Top Reads

0. A paper (first preprinted last year) by Bruno's Lab in collaboration with Baker's, Ovchinnikov's, and Vecchio's Labs, was published in Nature. The paper describes AF2seq-MPNN, a protocol/pipeline for the computational design of topologically complex protein folds and soluble analogues of membrane proteins. The pipeline integrates AF2seq, previously developed by Bruno's group, with ProteinMPNN from Baker's Lab. The approach was used to design soluble proteins that exhibit structural features of non-soluble GPCRs with high accuracy. Using motifs from the natural proteins, the designed analogues were functionalized. Read more...

1. A paper from Derek Woolfson's Group describes efficient rational/knowledge-guided approaches to designing small-molecule-binding barrel proteins. The paper, first preprinted last year, demonstrated the use of known sequence-to-structure relationships of α-helical peptide assemblies for rapid and accurate design of novel barrel-like proteins with accessible and functionalizable central channels. The approaches start from established parallel and antiparallel peptide assemblies to design two families of α-helical barrel proteins with inner channels that bind small molecules. Read more...

4. Bonus

I made a list of top 60 structural bioinformatics tools. Check it out here!

That's it for this week. Feel free to reach out by replying to this email if you have any suggestions or would like to have other discussions.

I try my best to make the information on this website as accurate as possible. If you find any errors in the contents of this page or any other page on this website, I would greatly appreciate that you kindly get in touch with me at contact@abeebyekeen.com.

If you are interested in joining my free weekly “BioModes and Top Reads” newsletter, please subscribe below.