MADI works by first breaking the initial protein sequence into fragments. These fragments are then manipulated using MADI’s AI and molecular dynamics simulation.
Through this process, MADI can predict the natural folding pattern for each protein segment as well as the desired folding pattern. Generating ten variations per hour, MADI is then able to refine candidates for optimal performance using stored and real-time data collected through MADI’s deep learning algorithm until the desired end-result.
MADI generates 10 candidate proteins per hour using scoring based on a docking model. This allows generation of cluster with candidates that demonstrate better performance and ultimately the best end result.
MADI is able to find relationships between sequence and structure. Each dot represents a given protein sequence with a given 3D structure.