Computational method to predict the solubility of proteins

Using a computational approach a team of scientists, led by Professor Michele Vendruscolo of the University of Cambridge, has developed a neural network method that can predict the solubility of a protein from the amino acid sequence and propose specific amino acid substitutions and/or insertions which will alter the solubility of the protein, while preserving its structure and functionality. The output is a short list of mutational variants with predicted solubility, or aggregation propensity, better than that of the protein provided as input. This method allows rapid screening of tens of thousands of mutations decreasing the time, cost and risk associated with the selection and development of candidate therapeutics and is of particular relevance for the development of therapeutics for high concentration subcutaneous formulations.