Computational protein engineering, scoring methodology, and practical notes from the team. Written for researchers, not marketers.
A deep dive into the thermodynamic and structural factors that make ΔΔG prediction from single substitutions genuinely difficult — and how modern ML models are beginning to close the gap.
How computational pre-screening can collapse a typical wet-lab variant library from thousands of candidates to a focused panel of five, without sacrificing diversity in the sequence landscape.
Mutations that increase binding affinity often destabilize the protein fold. We explore why this tradeoff exists, how to detect it computationally, and what it means for antibody and enzyme optimization campaigns.
For every enzyme variant that makes it to a pilot reactor, dozens were synthesized, expressed, and discarded. We quantify the cost of that attrition and the compounding timeline tax it places on bioprocess development.
AlphaFold changed structural biology — but it was trained to predict the most likely fold, not to evaluate how a single-residue substitution shifts thermodynamic stability or binding free energy. We explain the gap and what fills it.
Saturation mutagenesis of six CDR positions creates 8 billion sequence combinations. A focused computational pipeline can identify the highest-confidence affinity-stabilizing substitutions and shrink that space to a testable panel.
High-temperature industrial processes demand enzymes that stay folded under conditions far outside their natural range. We walk through the mutation classes that predictably shift Tm upward and the scoring functions that rank them reliably.
