OpenProtein.AI Platform

The OpenProtein.AI Platform is a deep learning-powered protein engineering solution designed for researchers and biotechnology teams seeking to accelerate the discovery and optimization of functional protein sequences. By combining machine learning models, generative AI, and intuitive data management tools, the platform enables scientists to design superior protein variants without requiring specialized computational expertise.

At the core of the platform is a web application that brings state-of-the-art ML-driven protein engineering to users at every skill level. OpenProtein.AI mines natural sequence databases and learns from experimental mutagenesis data to iteratively improve the design process, enabling the creation of variants with significantly enhanced activity compared to standard directed mutagenesis approaches.

Machine Learning-Guided Mutagenesis

• Generate novel variant libraries and predict their success across multiple functions of interest
• Visualize mutagenesis data and train machine learning models tailored to specific functions
• Define design objectives and build optimized variant libraries
• Improve multiple protein properties simultaneously to reduce experimental iterations
• Benefit from cumulative learning — every subsequent round and project builds on previous experimental data

Sequence-to-Function Mapping

• Develop and deploy predictive models based on your own experimental data to forecast activity for any input sequence
• Map all single-site substitutions to identify linchpin locations for site-saturating mutagenesis
• Visualize functional predictions for all single-site substitutions
• Export amino acid distributions for degenerate and combinatorial variant libraries

PoET — Generative Protein Design

• PoET (Protein Evolutionary Transformer) is an autoregressive, retrieval-augmented, generative transformer protein language model
• Infers the underlying evolutionary process from a set of representative sequences, learning functional constraints on amino acid sequences
• Generates novel, functional, and diverse protein sequences de novo — no functional or structural data required
• Scores the fitness of arbitrary query sequences under the learned evolutionary process
• Explores the local fitness landscape given a parent sequence and ranks specific variants to design focused mutagenesis libraries
• Validated on 90 different deep mutational scanning datasets across a wide range of protein families, organisms of origin, properties of interest, and MSA depths
• Models substitutions, insertions, and deletions as well as single and higher-order variants
• Supports prompt customization to define evolutionary context, custom MSAs from any sequence database, and in-software homology level settings to adjust sequence diversity
• Returns results in minutes with export options in multiple formats; free for academic use

Variant Library Design Features

• Evolutionary sequence analysis and generative protein language models
• Identification of mutagenesis hot spots
• Design of combinatorial variant libraries optimized for multiple design objectives
• Single site substitution, deletion, and insertion analyses
• Creation of likelihood-activity relationship generative models
• Statistical coupling analysis to discover areas with high potential for epistasis
• Design of single or higher-order variants with enhanced activity

Data Management and Platform Capabilities

• Secure, centralized data repository for large mutagenesis datasets
• Advanced in-app data management to track the full mutagenesis process
• Streamlined research workflows with all data managed in one place
• Intuitive web app interface requiring no specialized computational skills

The OpenProtein.AI Platform is accessible as a web application and is available via early access. It is free for academic use for the PoET generative design module, making cutting-edge AI-driven protein engineering broadly accessible to both industry and academic researchers.