Blog

AI-Designed Peptides Outperform GLP-1 in Lab

67% of AI-designed peptides outperformed GLP-1 in binding assays vs. ~0.1% for traditional methods. Peptides with 10+ changes delivered the biggest gains showcasing the power of AI creativity.

AI designs protein with applications for heart disease

For the first time, AI has taken general instructions and generated a functional protein. The AI-designed protein experimentally reacts to ATP, offering potential applications in treating heart disease, stroke, and more.

AI-Generated Proteins Achieve 84% Expression Success in the Lab

Discover how MP4, our transformer-based foundation model, is revolutionizing protein design by achieving an unprecedented 84% lab expression rate—vastly surpassing the traditional 20–30%. Dive into our journey of generating thousands of novel proteins and see what’s next with the development of MP5.

Testing 96 Novel AI Proteins

Our initial tests involved generating thousands of unique protein sequences, diverging from the 96 million sequences in the 3.8 billion-year evolutionary database. We then selected 96 challenging sequences to test in the lab, designed specifically to lack robust natural references.

Proteome enlightenment: AI annotation for proteins with unknown function

Advances in genome sequencing have uncovered millions of protein sequences, but only a tiny fraction have known functions. ProtNLM, a powerful AI tool by Google Research, aims to bridge this gap by predicting and annotating functions for previously uncharacterized proteins. With ProtNLM’s fast, accurate predictions, researchers can now rapidly gain insights into unknown proteins, supporting drug discovery, molecular biology, and more.

AlphaFold2, AlphaFold-Multimer, AlphaFold3

AlphaFold3 extends its predecessors’ capabilities by predicting structures not only for proteins but also for DNA, RNA, and their interactions. It marks a breakthrough in biomolecular modeling but still struggles with dynamic proteins and complex environments. Experimental validation remains essential, as AlphaFold3’s predictions, while powerful, can misrepresent certain structures—reinforcing the need for a balanced approach that combines both computational and experimental data for accurate structural biology insights.

Hide and seek: structure similarities with FoldSeek

FoldSeek is revolutionizing protein analysis by rapidly clustering millions of protein structures, revealing conserved patterns and uncovering unknown structures. It combines speed and sensitivity, making structural comparisons up to 10,000 times faster than traditional tools—essential for advancing drug design and evolutionary research.

I have several protein structures – what to do and where to start?

Starting with multiple protein structures? Structural similarity can guide your drug design journey by revealing evolutionary relationships and potential shared functions among proteins. Techniques like structural alignment and clustering enable deeper insights, helping to spot functional motifs and organize complex data for targeted research.

DiffDock, but better

The new DiffDock-L improves drug docking accuracy by 50%, with smarter training and more diverse data, but caution is still advised: even with broader binding pose predictions, each result should be thoroughly validated before use in drug design.

DiffDock for drug discovery

Discover how DiffDock is reshaping drug discovery with AI by generating precise ligand poses for proteins. This innovative tool brings both potential and challenges—especially in accuracy—making it a powerful yet cautious choice for researchers.

Drug design for beginners

Unlocking the 3D structures of proteins is a daunting task for labs, but with AI tools like AlphaFold2, even beginners can dive into drug design. Learn how molecular modeling and cutting-edge algorithms are reshaping biological research.

To bind, or not to bind, that was the question

Visualization showing ligand-binding predictions for lysozyme protein from rainbow trout (PDB ID: 1LMP) using the AF2BIND model. The image highlights residues ranked by binding probability, with a color gradient from blue (high binding potential) to red (low binding potential), indicating the most suitable binding sites for ligand interaction.

AI generated unique protein sequences with tunable diversity

Our AI designed new proteins with controllable diversity

From noise to molecules

How modern scientists use AI to create new proteins

Functional Protein Sequence Design using Large Language Models

Architectural advances in machine learning have accelerated de-novo protein design, adapting large language models for protein sequences to capture structural motifs. Unsupervised training on extensive protein data enables diverse artificial sequence generation, though current models face challenges in controllability and novelty.

ProteinMPNN: Message Passing on Protein Structures

Inverse folding leverages structural information to decode protein sequences through a machine learning model. Key innovations in ProteinMPNN include advanced encoding of structural details, updated message-passing schemas, and an order-agnostic decoding approach, enhancing sequence recovery accuracy.

Gen-AI protein

Gen-AI protein with varying loop length, sequence, and structure. In the context of an antibody

Zinc Binding by AI

AI Protein Design – Magnesium Binding

AI Protein Design – Zinc Binding

At 310.AI we’re building a foundational protein machine learning model that aims to understand everything

Antibody 101

Part 1: Antibodies are lobsters Part 2: How antibodies can kill cancer Part 3: What holds an antibody together?

NVIDIA – GTC 2023 Keynote

Large language models for large molecules

How AI protein structure prediction accelerates protein design

The promise of computational protein design is to replace slow, expensive, resource-intensive experimental methods with

Powering Computational Biology with ML Models: An Introduction to Our Serving Suite

We are thrilled to introduce our Bio ML Serving Product, a comprehensive platform that delivers

Benchmarking Machine Learning Methods for Protein Folding: A Comparative Study of ESMFold, OmegaFold and AlphaFold

Introduction Protein folding is a complex process that is essential for life. Proteins are made

Mol.E aims to design protein sequences directly from function

Natural proteins are one of the fundamental building blocks of living organisms. For decades, scientists have been engineering proteins to improve or even radically change their functions.