Unveiling AiCEsingle: Precision Redefined with Structural Constraints

In this study, researchers introduced AiCEsingle, a specialized module engineered to predict high-fitness (HF) single amino acid substitutions with remarkable accuracy. By extensively sampling inverse folding models—AI systems that propose compatible amino acid sequences from protein 3D structures—and integrating structural constraints, AiCEsingle significantly boosts the precision of protein variant predictions.

Figure 1. AI Breakthrough Redefines the Future of Protein Engineering.

Benchmarking Performance and Expanding Capabilities with AiCE

When tested against 60 deep mutational scanning (DMS) datasets, AiCEsingle consistently outperformed other AI-based methods, with accuracy improvements ranging from 36% to 90%. Its superior performance extended to complex proteins and protein–nucleic acid complexes. Notably, the inclusion of structural constraints alone led to a 37% gain in predictive accuracy. Figure 1 shows AI Breakthrough Redefines the Future of Protein Engineering.

To overcome the challenge of negative epistatic interactions in multi-mutation scenarios, the researchers developed AiCEmulti. This companion module incorporates evolutionary coupling constraints, enabling accurate identification of multiple high-fitness mutations with minimal computational overhead. This enhancement broadens AiCE’s applicability across a wider range of protein engineering tasks.

Engineering the Next Generation of Functional Proteins

Using the AiCE framework, the team successfully evolved eight structurally and functionally diverse proteins—including deaminases, nuclear localization signals, nucleases, and reverse transcriptases. These innovations led to the development of next-generation base editors tailored for precision medicine and molecular breeding, such as:

• enABE8e – A cytosine base editor with a ~50% narrower editing window
• enSdd6-CBE – An adenine base editor with 1.3× higher fidelity
• enDdd1-DdCBE –A mitochondrial base editor with 13× greater activity

A Powerful, Interpretable Tool for Protein Engineering

AiCE offers a streamlined, efficient, and broadly applicable strategy for protein design. By harnessing and enhancing existing AI models through structural and evolutionary constraints, it not only improves prediction performance but also boosts interpretability—marking a significant leap forward in AI-driven protein engineering.

Cracking the Code of Protein Design with AI

Protein engineering aims to create proteins with enhanced or entirely new functions, but predicting which mutations will work remains a massive challenge. Traditional methods are time-consuming and often miss subtle structural interactions. Enter AiCEsingle, an AI-powered tool that predicts high-fitness single amino acid substitutions by deeply analyzing a protein's 3D structure. By leveraging inverse folding models—which predict compatible sequences for a given structure—and enforcing structural constraints, AiCEsingle dramatically boosts accuracy, outperforming existing tools by up to 90%.

Going Beyond Single Mutations — Tackling Complex Protein Interactions

When designing proteins with multiple mutations, scientists face a common obstacle: epistasis, where the combined effect of mutations is worse than expected. To address this, the researchers developed AiCEmulti, a companion module that integrates evolutionary coupling constraints. This allows the system to predict multiple beneficial mutations that work well together—without falling into the trap of negative epistasis. Remarkably, AiCEmulti achieves this while keeping computational demands low, making it suitable for real-world, large-scale protein design efforts.

Real-World Impact — From Lab Bench to Precision Medicine

The true potential of AiCE shines in practical applications. Using its framework, researchers successfully evolved eight proteins with diverse roles, such as reverse transcriptases and nucleases. These innovations led to the creation of advanced base editors, key tools in gene therapy and crop engineering. Highlights include:

• enABE8e – A cytosine base editor with tighter editing precision
• enSdd6-CBE –An adenine editor offering greater fidelity
• enDdd1-DdCBE – A mitochondrial editor with 13× more activity

Source: SciTECHDaily

Cite this article:

Priyadharshini S (2025), Revolutionary AI Tool Transforms the Future of Protein Engineering, AnaTechMaz, pp.751