Back to Computing Overview
S1

Protein Engineering

From sequence to mutant, 7-step workflow

Four-model enzyme activity cross-validation. Single-model error 30-50%, only trusted when all four agree. Structure prediction -> catalytic activity -> stability screening -> sequence design -> immunogenicity assessment, full-process dry-computing loop.

2-4 weeks100K-800K/projectCapability Coverage 95%

Pipeline Flow

End-to-end dry-computing loop

1
Structure Prediction
ESMFold / Boltz-2 / Protenix multi-model prediction of mutant protein 3D structure
2
Catalytic Activity Scanning
DLKcat predicts enzyme activity changes, four-model cross-validation consensus
3
Stability Screening
FoldX ΔΔG thermal stability assessment, eliminates unstable mutants
4
Sequence Design
ProteinMPNN inverse folding generates optimized sequences
5
Immunogenicity Assessment
MHCflurry predicts MHC binding affinity
6
Multi-model Cross-validation
Only trusted when all four models agree
7
Synthesis-ready Delivery
Output DNA sequence + structure files + validation report

Core Tools

ESMFoldBoltz-2ProtenixProteinMPNNDLKcatFoldXMHCflurry

Core Advantages

Four-model Cross-validation

Enzyme activity trusted only when all four models agree, avoiding single-model errors

End-to-end Loop

From structure prediction to synthesis-ready, 7-step full process

Built-in Immunogenicity

Immunogenicity assessed at design stage

Capability Coverage

95%

USER SCENE

How This Technology Serves Patients

B2B2C: Through hospitals/doctors, this pipeline reaches these end-user scenarios

Start Your Project Protein Engineering

Start with just 1 project, no team needed.