PhiSign

Leveraging deep learning-based algorithms, we seek to design proteins with desired function, to use them to address some of the most complex and interesting challenges in biology, and to transform healthcare.

---

How to design proteins using statistical potentials?

An amino acid sequence contains the evolutionary information that is necessary for folding protein into its unique native structure, and evolutionary interactions between amino acids specify rules in stabilizing folds and functions of proteins. Although it is fundamental to address what exact interactions are sufficient and necessary to produce structure and biochemical function, decoding them into physical rules is challenging because of the complexity of cooperative interactions. Here, we attempt to statistically define the rules for guiding rational protein design using evolutionary data derived from the multiple sequence alignment in absence of structural information.