Skip to main content

 9XZP | pdb_00009xzp

Crystal structure of AI-designed homodimer MIYEI


Experimental Data Snapshot

  • Method: X-RAY DIFFRACTION
  • Resolution: 1.77 Å
  • R-Value Free: 
    0.245 (Depositor), 0.257 (DCC) 
  • R-Value Work: 
    0.189 (Depositor), 0.198 (DCC) 

Starting Model: in silico
View more details

wwPDB Validation 3D Report Full Report

Validation slider image for 9XZP

This is version 1.3 of the entry. See complete history. 

Literature

An Artificial Intelligence Model for Translating Natural Language into Functional De Novo Proteins.

Riley, T.P., Parsa, M.S., Kalantari, P., Naderi, I., Azimian, K., Begloo, N., Matusovsky, O., Osher, E., Prabhakaran, D., Yadav, R., Simelis, K., Mathavan, I., Singh, J., Azimian, K., Wei, K.Y.

(2026) ACS Synth Biol 15: 3645-3652

  • DOI: https://doi.org/10.1021/acssynbio.6c00246
  • Primary Citation Related Structures: 
    9XZO, 9XZP

  • PubMed Abstract: 

    Traditional protein design is fundamentally constrained by known sequences and folds. To break free from these limitations, we introduce a new alternative: designing proteins directly from plain-language specifications. To achieve this, we trained MP4, a transformer-based model that maps natural language prompts to protein sequences, on a dataset of 3.2 billion points and 138k tokens. In a benchmark of 96 prompts representing a wide array of functions and contexts, MP4 excelled by simultaneously improving on three key metrics: sequence realism, predicted fold quality, and alignment to the requested function. This high performance is particularly significant as it was achieved using only text as input which is a major departure from other models. Experimental validation confirmed our computational predictions: two de novo designs were experimentally shown to be both expressible and thermostable, with high-resolution crystallography (1.30 Å and 1.77 Å) ultimately revealing one to possess a paradigm-shifting novel fold. Functionally, the designs were also active, demonstrating both ATP binding and hydrolysis in vitro. This work demonstrates the realization of natural-language intent as functional proteins that express, crystallize, and catalyze. Although the underlying approach is still in early development with incomplete coverage and controllability, MP4 delivers a profound impact: it lowers the barrier to protein design and vastly expands the space for creative exploration in molecular programming.


  • Organizational Affiliation: 
    • 310 AI, San Francisco, California94107, United States.

Macromolecule Content 

  • Total Structure Weight: 16.06 kDa 
  • Atom Count: 1,298 
  • Modeled Residue Count: 155 
  • Deposited Residue Count: 156 
  • Unique protein chains: 1

Macromolecules

Find similar proteins by:|  3D Structure
Entity ID: 1
MoleculeChains  Sequence LengthOrganismDetailsImage
De novo protein
A, B
78synthetic constructMutation(s): 0 
Entity Groups
Sequence Clusters30% Identity50% Identity70% Identity90% Identity95% Identity100% Identity
Sequence Annotations
Expand
Reference Sequence

Experimental Data & Validation

Experimental Data

  • Method: X-RAY DIFFRACTION
  • Resolution: 1.77 Å
  • R-Value Free:  0.245 (Depositor), 0.257 (DCC) 
  • R-Value Work:  0.189 (Depositor), 0.198 (DCC) 
Space Group: C 1 2 1
Unit Cell:
Length ( Å )Angle ( ˚ )
a = 53.093α = 90
b = 53.473β = 112.443
c = 55.786γ = 90
Software Package:
Software NamePurpose
autoPROCdata processing
Aimlessdata scaling
TRUNCATEdata processing
REFMACrefinement
XDSdata reduction
PHASERphasing

Structure Validation

View Full Validation Report



Entry History 

& Funding Information

Deposition Data

  • Released Date: 2025-10-22 
  • Deposition Author(s): Wei, K.Y.

Funding OrganizationLocationGrant Number
Not funded--

Revision History  (Full details and data files)

  • Version 1.0: 2025-10-22
    Type: Initial release
  • Version 1.1: 2026-04-15
    Changes: Database references
  • Version 1.2: 2026-09-02
    Changes: Database references
  • Version 1.3: 2026-11-04
    Changes: Database references