Skip to main content

​
The Problem You Are Trying to Solve

“I have a set of lead compounds, and I want to optimize them across multiple developability dimensions, like solubility, ADME, toxicity, and related properties, without losing on-target potency.”
Hero Image
At the lead optimization stage, teams often face tradeoffs:
  • A potent binder may be insoluble or poorly permeable
  • A promising series may carry CYP liabilities, hERG risk, or DILI flags
  • Small structural changes can meaningfully shift clearance, bioavailability, or safety
Multi-parameter optimization (MPO) is about balancing the full profile, not maximizing a single metric.

​
Solution

This workflow uses a primary MPO loop that alternates between:
  1. Designing new analogs with the right constraints
  2. Scoring them across developability endpoints
  3. Filtering for synthesize-ability and feasibility
  4. Repeating until a balanced set emerges
The primary MPO chain on Revilico is: Baseline Profiling → Property Screening (Solubility + ADMET-AI) → Multi-Objective Molecular Optimization → Synthesis Feasibility (Retrosynthesis) → Iterate. Binding engines can be integrated at any stage to ensure potency and selectivity are not sacrificed while optimizing developability.

​
What Data Do I Need to Provide?

Required
  • Lead structures (SMILES / CSV)
Recommended
  • Your desired optimization targets (examples: “increase solubility, reduce CYP3A4 inhibition risk, maintain MW < 500”)
  • Optional hard constraints (avoid specific motifs, preserve pharmacophore features, keep series scaffold)
Optional
  • Experimental solubility/clearance/tox data (if available)
  • On-target structure/assay signals for potency anchoring

​
Workflow

  1. Generate New Analogs Using Multi-Objective Molecular Optimization
Design new molecules that improve multiple endpoints simultaneously. Use Generative Chemistry → Molecular Optimization to:
  • Generate analogs from the existing leads (conservative or exploratory)
  • Ensure that there is proper activity of your ligands to the target of interest during this process
  • Encode MPO goals as a multi-component scoring function (e.g., solubility ↑, CYP risk ↓, hERG risk ↓, MW/LogP constraints, etc.)
  • Apply diversity controls so you don’t get 1,000 near-duplicates
This step is the core MPO engine: sample → score → update → repeat, producing candidates increasingly aligned to your objective profile, resulting in a versioned library of optimized analogs ranked by multi-objective score
  1. Establish the MPO Baseline
After generating new molecules, define the success criteria for the lead series. On Revilico, users typically:
  • Run ADMET-AI Analysis to surface the major liabilities (toxicity flags, CYP inhibition/substrate risks, clearance risk, permeability proxies)
  • Run Compound Solubility to understand baseline solubility trends across solvents and temperature
  • You can run the Ligand Membrane MD Engine that will allow you to analyze permeability of the compound to different seeded cell membranes
  • (Optional) Run Conformer Search or other quantum methods to understand flexibility and shape drivers that may impact permeability and exposure
This baseline informs:
  • Which liabilities matter most
  • Which constraints must remain fixed
  • How aggressive optimization should be
This step provides a clear MPO objective (prioritized endpoints + acceptable ranges).
  1. Rapid Property Screening and Downselection
Run fast, high-throughput filters to avoid spending time optimizing the wrong chemistry. Use:
  • ADMET-AI Analysis to identify red flags early (toxicity, CYP panels, hERG/DILI risk, clearance risk, permeability)
  • Compound Solubility to identify which leads have a solubility ceiling that must be addressed by design
At this stage, you are not looking for perfection, you are identifying:
  • which leads are salvageable
  • which liabilities are dominant
  • which properties move together vs trade off
This gives a ranked lead shortlist + a prioritized “liability map” per compound.
  1. Interpret and Filter Results
After the optimization run, identify candidates that are balanced, not just “good at one metric.” Common selection heuristics:
  • Prefer compounds with consistent ADMET improvements (not one extreme gain paired with new liabilities)
  • Avoid candidates that only score well due to compensation effects
  • Maintain chemical diversity among finalists to reduce risk of series-level blind spots
Use:
  • ADMET-AI Analysis (re-run on the new library) for confirmation
  • Compound Solubility (re-run for top candidates) to validate solubility gains
This produces a short list of MPO-balanced candidates for synthesis planning
  1. Check Conformations, Electronic Risk, and Physical Plausibility (Optional)
For finalists, deepen confidence before synthesis. Use:
  • Conformer Search to ensure the compounds can adopt relevant 3D shapes without extreme strain
  • Geometry Minimization and Thermochemistry to confirm stable geometries and rule out unstable/high-energy structures
  • Molecular Orbital Analysis (HOMO–LUMO) for electronic reactivity signals (useful for flagging unstable/reactive chemistry)
This step supports “sanity checks” that catch hidden risks, giving physics-backed confirmation that finalists are structurally and electronically reasonable.
  1. Ensure Synthesizability with Retrosynthesis
A candidate that cannot be made is not a lead. Use Retrosynthesis to:
  • Propose ranked synthetic routes
  • Flag candidates with unrealistic or costly synthesis pathways
  • Prioritize compounds that are both better and buildable
This engine results in a buildable shortlist with actionable synthetic plans.

​
Results

  • A refined set of candidates improved across solubility, ADME, and toxicity dimensions
  • Explicit tradeoff awareness (what improved, what worsened, and why)
  • Higher confidence in which compounds are worth synthesis and assays
  • A repeatable MPO loop you can run iteratively as new data arrives

​
Integration with Other Engines (Optional)

If potency/selectivity must be enforced within the MPO loop, users can integrate:
  • Docking / Virtual Screening to maintain binding hypotheses while improving developability
  • Protein–Ligand MD to validate binding stability for optimized candidates
  • ABFE/RBFE for high-confidence ranking among close analogs
This allows MPO to be developability-first without losing efficacy.

​
Why Revilico?

Revilico enables MPO as a single connected loop:
  • Generative design (Molecular Optimization) grounded in explicit scoring
  • Fast developability screens (ADMET-AI, Solubility) to guide iteration
  • Physical plausibility checks (Conformer Search, QM engines) for confidence
  • Buildability validation (Retrosynthesis) to ensure real-world feasibility
This produces candidates that aren’t just more potent, but actually more likely to succeed in development.