Machine Learning for Guided Fluorescent Protein Evolution
Machine Learning for Guided Fluorescent Protein Evolution
Embedded Files
Supervised Learning from Large Mutational Datasets
Supervised Learning from Large Mutational Datasets
Vectorized Sequence Data Can Transform Complex Amino Acid Composition into Useful and Predictive Inputs for Model Training that Enable In-Silico Design and Wet-Bench Workflows.
Deep Mutation of Sequences Reveals Sites for Adjusting Key Parameters
In-Silico Deep Mutational Scanning and Automated Data Mining Enable Accurate Residue-Level Predictions
Implementing Greedy Search Enables Molecule Design Across Parameters
Implementing Greedy Search Enables Molecule Design Across Parameters
De Novo Fluorescent Protein Design Using Protein Language Models
De Novo Fluorescent Protein Design Using Protein Language Models
Model Trained on FPbase.org Sequences
pLM Metrics
Smooth Loss Convergence
Perplexity Reduction to 3 Amino Acids
Stable Cosine LR Decay
Tunable to Lineage Specific Datasets
Structural Evaluation of pLM-Generated Designs
Example Design 1
Example Design 2
Example Design 3
High Entropy Example Design
Sample MSA of LM Generated Designs (30-50% Sequence Identity)
Note: Design Can Reach 60-80% Sequence Identity by Filtering FP Lineage
Assisting Directed Evolution Campaigns - It's Super Effective
Our platform supercharges your directed evolution campaigns by combining traditional experimental strategies with modern machine learning tools. Just like a strategic move in a classic game, our guided approach helps you identify the most promising mutations faster, reduce screening burdens, and achieve brighter, more functional fluorescent proteins. It’s not just effective — it’s super effective.
Let us help you level up your protein engineering game.
Page updated
Google Sites
Report abuse