WiseOmni Blog

Sharing AIDD AI agent technology, industry insights, and latest updates

Chai-2: The "24-Well Plate" Revolution in Antibody Design

Chai-2 is a multimodal generative model achieving 16% experimental success rate in de novo antibody design, representing a 100x improvement over previous methods, reducing discovery timeline from months to two weeks.

Read more →

Chai-1: Advances in Multimodal Molecular Structure Prediction

Chai-1 is a multimodal molecular structure prediction foundation model achieving state-of-the-art performance in protein-ligand interaction and protein multimer prediction, supporting experimental constraint prompting and single-sequence prediction.

Read more →

RFdiffusion: When Diffusion Models Meet Protein Design

RFdiffusion represents a significant paradigm shift in protein design—introducing diffusion models from image generation to protein structure generation, achieving remarkable advances in unconditional monomer design, protein binder design, and symmetric oligomer design.

Read more →

OpenFold3 Technical Assessment: Performance and Limitations

OpenFold3-preview2 is the latest open-source reproduction of AlphaFold3, developed by Columbia University and Lawrence Livermore National Laboratory. It is the only academic reproduction supporting training from scratch.

Read more →

OpenFold Technical Assessment: Training Mechanisms and Generalization Analysis

OpenFold is a complete open-source reproduction of AlphaFold2 developed by Columbia University and others, with full training code, model weights, and datasets publicly available.

Read more →

Evo 2: Genome Modeling and Design Across All Domains of Life

Evo 2 is a biological foundation model by Arc Institute, Stanford, and NVIDIA, trained on 9.3 trillion DNA base pairs covering all domains of life, achieving 1 million token context window and single-nucleotide resolution.

Read more →

Evo: Multimodal Biological Sequence Modeling from Molecular to Genome Scale

Evo is a 7B parameter genomic foundation model developed by Arc Institute and Stanford University, using StripedHyena architecture for single-nucleotide resolution long-sequence modeling with zero-shot functional prediction across DNA, RNA, and protein modalities.

Read more →

Genomic Language Models: Opportunities and Challenges in Cross-Scale Modeling

Genomic Language Models (gLMs) represent an emerging field applying NLP techniques to DNA sequence analysis, showing potential in functional constraint prediction, sequence design, and transfer learning while facing unique challenges of genome scale and sparse functional regions.

Read more →

ESMFold Technical Analysis: Language Model-Driven Single-Sequence Protein Structure Prediction

ESMFold is Meta AI's protein language model-based single-sequence structure prediction method, achieving AlphaFold2-comparable accuracy without MSA, with up to 60x speed improvement.

Read more →

Protein Language Models: Technical Evolution, Core Challenges, and Future Directions

A comprehensive review of Protein Language Models (PLMs), covering architectural evolution, positional encoding strategies, scaling laws, dataset construction, and downstream applications, including ESM series, ProGen2, and analysis of MSA-free and multi-modal fusion trends.

Read more →

VirSentAI: Autonomous Multimodal Agent for Zoonotic Surveillance and Drug Repurposing

VirSentAI is an autonomous trimodal agent developed by University of A Coruña, integrating MedGemma, HyenaDNA, and PLAPT models for viral surveillance and drug repurposing. Achieves AUROC 0.95 across 31,728 complete viral genomes.

Read More →

Fleming: An Integrated AI Agent for Tuberculosis Antibiotic Design

Fleming is an integrated AI agent for TB antibiotic discovery developed by Harvard and other institutions. Achieved 83% hit rate in prospective validation of 435 molecules, with 100% hit rate for 6 de novo designed molecules in wet-lab validation.

Read More →

Latent-Y: Technical Assessment of a Lab-Validated Autonomous Drug Design Agent

Latent-Y is the first AI agent capable of autonomous end-to-end biologics design from natural language prompts. In wet-lab validation across 9 targets, 67% success rate with 56x efficiency improvement, marking a paradigm shift from molecular design to scalable execution.

Read More →

The Social Turn of Agentic AI: Re-examining the "Intelligence Explosion" Narrative

Based on Science commentary: Intelligence is essentially a high-dimensional, relational social attribute. Research shows reasoning models improve accuracy through multi-perspective debates within an internal "society of thought."

Read More →

BoltzGen: Universal Binder Design with All-Atom Generative Models

BoltzGen is an open-source all-atom generative model developed by MIT, Valence Labs, and others for designing protein and peptide binders, achieving 66% nM-level success rate across 8 wet-lab validation projects.

Read More →

Boltz-2: Unified Framework for Structure and Affinity Prediction

Boltz-2 is an open-source biomolecular structure prediction model developed by MIT, Valence Labs, and ETH Zurich, achieving AI binding affinity prediction accuracy approaching FEP methods with 1000x+ speedup.

Read More →

Boltz-1: Open-Source Biomolecular Interaction Prediction

Boltz-1 is an open-source biomolecular structure prediction model developed by MIT, Genesis Research, and CHARM Therapeutics, achieving multiple innovations on AlphaFold 3 architecture with comparable prediction accuracy.

Read More →

IsoDDE: Generalization Leap in Biomolecular Interaction Prediction

IsoDDE is a drug design engine developed by Isomorphic Labs, achieving significant improvements over AlphaFold 3 architecture, with substantial performance gains in protein-ligand and antibody-antigen interface prediction.

Read More →

AlphaFold 3 Technical Report Analysis

AlphaFold 3 is a biomolecular structure prediction model developed by DeepMind and Isomorphic Labs, using diffusion-based architecture to uniformly predict complex structures including proteins, nucleic acids, small molecules, ions, and modified residues.

Read More →

AlphaFold2 Protein Structure Prediction: In-Depth Analysis

AlphaFold2 represents a milestone breakthrough in protein structure prediction, achieving experimental-level accuracy for the first time. This article provides an in-depth analysis of its technical principles, performance, and applications.

Read More →

AIDD Agents: The Dawn of a New Era

The arrival of the AIDD agent era marks the transformation of AI drug discovery from specialized tools to infrastructure. This article predicts six major changes including increased base model usage, autonomous workflow planning, and DMTA cycle optimization.

Read More →