Skip to main content
The Problem You are Trying to Solve:
“I have identified a molecular target implicated in disease, and I want to determine the optimal therapeutic strategy/ mode of action (e.g., inhibition, activation, or allosteric modulation) to achieve the desired biological outcome”
Hero Image
Traditionally, determine optimal therapeutic strategy required extensive experimental validation through costly and time-consuming assays: enzymatic activity screens to test inhibition, cell-based phenotypic assays to assess pathway modulation, co-immunoprecipitation experiments to validate protein-protein interactions, and often clinical trial data to reveal that the wrong modality was chosen after years of development. The disconnect between target biology, target structure, and cellular context meant therapeutic strategy selection was based on precedent rather than comprehensive mechanistic understanding. Now with AI-powered structural prediction, transcriptomic profiling, and computational binding analysis, we can integrate target structure, biological function, and systems level pathway effects to rationally predict optimal therapeutic approaches before committing to expensive experimental campaigns Solution
This workflow enables users to a) systematically evaluate all druggable sites on a target protein through structural pocket analysis and b) determine optimal therapeutic strategy and mode of action by integrating structural druggability, transcriptomic pathway analysis, and dynamic binding site characterization across Revilico’s multi-modal engine suite. We leverage complementary computational methods such as pocket identification for binding site discovery, scRNA-seq analysis for biological context and pathway validation, molecular dynamics for conformational flexibility assessment, and co-folding predictions for protein-protein interface analysis, to ensure therapeutic strategy selection is grounded in both structural feasibility and system-level biological understanding What Data Do I Need to Provide?
  • Protein sequence (used to generate the different possible protein structures with Alphafold)
Workflow
  1. Generate or Obtain Target Protein Structure
Establish a high-confidence 3D structural model of your target protein to enable pocket identification and druggability assessment. We will first use RevilicoGPT/Revilico Agent with the following query: “I am evaluating AXL for Triple Negative Breast Cancer. Search PDB and retrieve any experimental structures, prioritizing high-resolution structures (< 2.5 A) with bound ligands or in different functional states (apo, holo, active, inactive).” If there are no suitable experimental structures, we can run Alphafold or Openfold using our protein sequence, where we will prioritize structures with high confidence scores. Usually, our team likes to take co-crystal structures of known inhibitors with ideal modes of action to derive pocket coordinates and to understand ideal amino acids driving binding within key regions in the pocket.
  1. Identify Druggable Binding Sites
We will run Pocket Search Engine on our protein structure to identify all cavities and rank by druggability score, volume, and physicochemical properties. Note pocket location, i.e. catalytic site pockets suggest orthosteric inhibition, surface pockets away from active sites suggest allosteric modulation, and float protein-protein interfaces suggest PPI disruption strategies. For PPI targets, use BoltzGen Co-Folding to model the protein complex interface. Analyze interface area, binding energy, and hot-spot residues. Weak interfaces with shallow binding sites suggest PPI disruption is feasible; strong interfaces with deep binding grooves may require stabilization, alternative therapeutic modalities, or indirect modulation strategies.
  1. Analyze Biological Context Via Transcriptomics
Use Revilico Agent to obtain relevant scRNA-seq datasets comparing disease vs control or perturbed vs baseline states for your target. Here is a sample query: “I am studying EGFR in Triple Negative Breast Cancer. Find scRNA-seq datasets in h5ad format comparing disease vs normal. The datasets must include EGFR and be from the cell line MDA-MB-231.” You will then run DEG Analysis and Automated Target ID from scRNA-seq Analysis. From the results we can extract the DEG volcano plot from DEG Analysis to see how differentially expressed our target gene is from control and experimental. From the Automated Target ID section we can see if downstream genes from our target are affected. We can also reference the phenotype bar graph to validate if our phenotype predictions of the target align with it. If we are seeing an over-activation of certain genes, we can use this information to design a mode of action that inhibits the translated protein function, therefore returning the system back to equilibrium.
  1. Assess Conformational Flexibility
Run Protein Water MD simulations followed by MDPocket analysis to identify cryptic or transient products that only appear during protein motion. Cryptic pockets that open frequently represent opportunities for allosteric modulation that static analysis would miss. This extracts multiple protein conformations from the trajectory and shows which conformational states are thermally accessible. It will also reveal whether the protein is rigid or highly flexible.
  1. Validate Strategy with Proof-of-Concept Docking
For orthosteric inhibition run Static Docking or Flexible Docking with known substrates, cofactors, or reference inhibitors against the active site pocket. Strong binding scores with chemically reasonable poses validate that the small molecule can effectively compete with natural substrates. For allosteric modulation, dock small fragment libraries or known allosteric modulators into cryptic//allosteric pockets identified. Successful binding to these sites with stable poses suggest allosteric strategy is viable. Run Protein-Ligand MD on top poses to confirm allosteric pockets remain stable when occupied, and that they make conformational modifications to the entire protein structure (especially within the active site). This can be confirmed by looking at larger jumps in Root Mean Squared Fluctuatio (RMSF) within pocket residues. For PPI Disruption, use Boltzgen CoFolding to model binding of peptide mimetics or small molecules at the protein-protein interface. Alternatively, dock fragment libraries targeting interface hot-spots. Validate using Pharmacophore Analysis that compounds can recapitalize key interface interactions and engage with key amino acids driving biological activity. Results
  • Ranked druggable pockets with druggability scores, volumes, and strategic classifications
  • Transcriptomic response profiles
  • Conformational flexibility assessment
  • Proof-of-concept binding validation
  • Recommended therapeutic strategy with structural and biological justification
Convergent evidence across structural druggability, biological phenotypes, and binding validation determines optimal strategy. For orthosteric inhibition: look for high scoring active site pockets, strong transcriptomic response showing target perturbation drives desired phenotypes, and validated substrate competitive binding. For allosteric modulation: prioritize cryptic pockets appearing >30% of MD simulation, moderate transcriptomic effects suggesting partial modulation suffices, and stable allosteric ligand binding in validation docking. For PPI disruption: weak interfaces, phenotypic predictions indicating complex formation drives pathology, and successful peptide/fragment binding at interface hot-spots. Discordant results suggest the target may require alternative modalities like protein degradation or may not be optimally druggable, prompting reconsideration of target selection or combination strategies. Now what? After determining your mode of action and biophysical function of your system, you can move into iterative compound design.
  • Utilizing all of the engines within the Revilico platform from binding chemistry to generative chemistry, you can run a variety of assessments on your compounds for further synthesis and testing.
Why Revilico?
This workflow determines the optimal therapeutic strategy for a molecular target (e.g., inhibition, activation, or PPI disruption) by integrating several key data points. This involves assessing the target’s structural druggability, confirming that perturbation drives the desired biological phenotype (transcriptomics), and validating a stable binding mode with compound studies.