The Chai discovery team comes out chai-2: the AI ​​model reached a 16% success rate in the design of Novo antibodies

TLDR: Chai Discovery Team introduces Chai-2, a multimodal AI model that enables zero-shot de novo antibody design. Achieving a 16% hit rate across 52 novel targets using ≤20 candidates per target, Chai-2 outperforms prior methods by over 100x and delivers validated binders in under two weeks—eliminating the need for large-scale screening.

In an important progression for the discovery of computer medications, the Chai discovery team introduced Chai-2A multimodal generative AI platform capable of zero antibodies and a protein binder design. Unlike previous approaches based on high-speed extensive screening, CHAI-2 reliably designs functional binders in a single 24 well Configuration, realizing More than 100 times improvement On existing cutting -edge methods (SOTA).

Chai-2 was tested on 52 new targetsNone of which had experienced antibodies or nanobodons binders in the protein data bank (PDB). Despite this challenge, the system has reached a 16% experimental success rateDiscover binders for 50% of the targets tested in a two -week cycle From computer design to the validation of the wet laboratory. This performance marks a transition from probabilistic screening to the deterministic generation in molecular engineering.

Novo design on an experimental scale on an experimental scale on a experimental scale

Chai-2 incorporates a Generative design module of all atom and a folding model that predicts complex antibody-antigen structures with double the precision of its predecessor, chai-1. The system works in a Zero-shot adjustmentgenerating sequences for antibody methods such as SCFV and VHH without requiring previous binders.

The key characteristics of Chai-2 include:

• No specific target adjustment required
• Ability to Quick conceptions using constraints in the epitope
• Generation of Therapeutically relevant formats (miniproteins, scfvs, VHHS)
• Sustain Cross reactivity design Between species (for example, human and cyno)

This approach allows researchers to design ≤20 antibodies or target nanobodies and completely bypass the need for high speed screening.

Benchmarking through various protein targets

In rigorous laboratory validations, the chai-2 was applied to the targets with No sequence or structure similarity with known antibodies. The conceptions were synthesized and tested using Bio-layer interferometry (BLI) For the connection. The results show:

• Average success rate of 15.5% In all formats
• 20.0% for VHH,, 13.7% for SCFV
• Successful binders for 26 of the 52 targets

In particular, Chai-2 produced hits for hard targets such as TNFαwhich was historically insoluble for the in silico design. Many binders have shown Constant pecheomolar dissociation with low nanomolar (KDS)indicating high affinity interactions.

New, diversity and specificity

CHAI-2 outputs are structurally and sequentially distinct from known antibodies. Structural analysis has shown:

• No design generated had <2å rmsd of a known structure
• All CDR sequences had a modification distance> 10 of the nearest known antibody
• The binders fell into several structural clusters by target, suggesting conformational diversity

Additional assessments confirmed Outside Lower Liaison And Comparable polyreacactivity profiles clinical antibodies such as trastuzumab and iXizumab.

Flexibility and design customization

Beyond the generation of general use, chai-2 demonstrates the capacity of:

• Multiple target epitopes on a single protein
• Produce binders through Different antibody formats (for example, SCFV, VHH)
• Generate Inter-species reactive antibodies in an invite

In a case-reactivity study, an antibody designed by CHAI-2 has reached Kds nanomolar Against the human and cyno variants of a protein, demonstrating its usefulness for Preclinical studies and therapeutic development.

Implications for the discovery of drugs

Chai-2 effectively compresses the traditional calendar of the biological discovery of month to weeksOffering wires validated experimentally in a single turn. Its high success rate combination, new design and modular incentive marks a paradigm change in therapeutic discovery workflows.

The frame can be extended beyond antibodies to Miniproteins, macrocycles, enzymesand potentially small moleculespaving the way to Computational design paradigms in computer science. Future orientations include expansion Bispecific, ADCand explore Optimization of biophysical property (for example, viscosity, aggregation).

As the field of AI in the ripened molecular design, Chai-2 defines a new bar for what can be made with generative models in real drug discovery parameters.

---

Discover the Technical report. All the merit of this research goes to researchers in this project. Also, don't hesitate to follow us Twitter,, YouTube And Spotify And don't forget to join our Subseubdredit 100k + ml and subscribe to Our newsletter.