Molecule CoPilot
CoPilot Lead Optimizer is an AI assistant for drug discovery, offering on-the-fly lead optimization based on user prompts. Whether enhancing drug-likeness or adjusting properties like synthetic accessibility, CoPilot excels using just the ligand SMILES and user input. It can also utilise the protein-ligand complex to optimize protein-ligand binding, generate lead-like molecules, provide real-time docking scores for newly generated molecules, and divulge the scientific reasoning behind its decisions.

📄️ User Inputs
The simplest input to CoPilot is the ligand smiles along with a user-specific prompt (under the fields “Enter a SMILES String”, “Enter the desired Modifications” in Figure 1, respectively), this is particularly aimed at property optimisation and generates new molecules with improved target properties. The user provides a single input SMILES string. Pasting a correct SMILES string should produce a 2D image of the molecule, on the other hand, an incorrect SMILES will generate an error “Invalid SMILES string. Please enter a valid SMILES.” Along with the SMILES, the user is required to provide a prompt describing the desired modification for the molecule. These modifications can be broad, such as improving QED, or highly specific, such as modifying a specific chemical group on the molecule. Upon clicking “Generate Molecules”, CoPilot provides the user with five molecules with the desired modifications as requested by the user.
📄️ Job Submission
The user may submit the job by clicking on the “Submit” button. CoPilot is able to derive the results in real time and within less than a minute provides 5 output results in the form of a collapsable window.
📄️ Results
Figure 2\. The Modifications section from CoPilot displays the 5 outputs
📄️ Further Modifications
Co-pilot allows the user to further customise their lead optimisation process by modifying any newly generated molecules. This allows the user to incorporate the often iterative drug discovery process and make improvements/modifications along the way.