Crystal Landscape Exploration
Reduce the risk that a plausible low-energy packing arrangement remains outside the experimental plan.
- Conformer and packing enumeration
- Space-group-aware candidate generation
- Duplicate and motif clustering
AI for Pharmaceutical Solid Form Development
AI for Polymorph Screening
Map the solid-form landscape before an unseen crystal form becomes a late-stage development problem. CD ComputaBio combines AI-assisted crystal landscape exploration, physics-based ranking, and evidence-led experimental planning to prioritize the forms and conditions that matter.
Landscape firstExpose plausible low-energy forms
Risk focusedPrioritize what could disrupt development
Experiment linkedConvert predictions into testable conditions
Service coverage
Polymorph Screening Services
AI-Assisted Candidate Triage
Concentrate high-cost calculations and laboratory effort on candidates most likely to alter a development decision.
- Surrogate and ML force-field ranking
- Structural diversity selection
- Uncertainty-aware shortlisting
Relative Stability Assessment
Distinguish low-energy structures from forms that remain competitive under relevant temperature and pressure conditions.
- Periodic energy refinement
- Free-energy and phonon options
- Thermodynamic ranking
Targeted Crystallization Design
Translate landscape gaps into solvent, thermal, slurry, seeding, cooling, evaporation, and pressure experiments with clear hypotheses.
- Condition-to-form rationale
- High-information experiment selection
- Iterative result feedback
Polymorph Risk & Control Strategy
Define what evidence supports progression—and which transformation pathways should remain under surveillance.
- Form relationship map
- Conversion risk scenarios
- Decision-ready technical dossier
Polymorph risk atlas
Polymorph Risk Assessment
We evaluate energy, structural novelty, kinetic accessibility, environmental sensitivity, and evidence gaps together. The result is a transparent priority map rather than a single “best-form” prediction.
Energy proximity
Packing novelty
Condition access
Conversion risk
Evidence gap
Lower attentionIllustrative prioritizationHigher attention
Adaptive strategy
Polymorph Screening Workflow
01
Frame the form risk
Align the screen with route, formulation, process exposure, known forms, analytical evidence, and the decision the program must support.
02
Expand plausible packing
Sample conformers and packing arrangements, then cluster redundant structures to preserve meaningful landscape diversity.
03
Refine competitive forms
Apply progressively higher-fidelity ranking—from efficient potentials to ML force fields and periodic quantum methods where warranted.
04
Challenge with conditions
Evaluate temperature, pressure, hydration or solvation scenarios and design experiments capable of resolving the highest-impact uncertainties.
05
Close the evidence loop
Reconcile PXRD, thermal, spectroscopic, microscopy, or crystallization observations with predicted structures and update the control strategy.
Project readiness
Project Inputs and Deliverables
Recommended inputs
- 2D/3D API structure, protonation state, stereochemistry
- Known forms, solvates, hydrates, salts, or amorphous observations
- PXRD, DSC/TGA, spectroscopy, microscopy, and solubility data
- Crystallization conditions, process history, and form-selection target
Quality controls
- Traceable assumptions and versioned candidate sets
- Energy-method escalation matched to decision risk
- Structural deduplication and motif review
- Explicit uncertainty, limitations, and experimental validation needs
01Ranked crystal landscape
02Structural families & motifs
03Targeted experiment matrix
04Polymorph risk dossier
Published data
CASE 01 · RATIONAL EXPERIMENT DESIGN
CASE 02 · HIERARCHICAL AI + PHYSICS
Published Polymorph Studies
The schematics below are original method-to-decision summaries created for this page; they do not reproduce publication figures or represent CD ComputaBio project results.
Prediction directed a high-pressure crystallization test
LandscapeGenerate and rank plausible crystal structures for a flexible pharmaceutical.
→
Condition hypothesisUse pressure-dependent stability to identify a route to a missing dense form.
→
Decision evidenceTarget experiments and reassess the risk of a late-appearing stable polymorph.
Flexible APIEnergy landscapePressure perturbationForm-risk reduction
Neumann and colleagues combined crystal structure prediction with high-pressure crystallization for dalcetrapib, demonstrating how an in silico landscape can propose a focused experimental condition and help assess hidden-form risk.[1]
View publicationML force fields accelerated large-scale candidate ranking
Packing searchSystematically explore symmetry and packing variables across candidate structures.
→
Tiered rankingFilter with molecular mechanics, machine-learning force fields, and periodic DFT.
→
Development useShortlist known and potential low-energy forms for experimental and formulation planning.
Systematic searchMLFF triagePeriodic DFTCandidate shortlist
Zhou and colleagues reported a hierarchical CSP workflow validated on 66 molecules and 137 experimentally known polymorphic forms, illustrating how ML-assisted ranking can extend the reach of physics-based screening.[2]
View publicationProject decision guide
Polymorph Screening FAQs
Scope and fidelity should follow the development decision, molecular complexity, and evidence already available.
Can the project start before experimental polymorphs are known?
Yes. A structure-led landscape can identify plausible packing families and help design an initial, information-rich experimental screen. Results are treated as hypotheses that require analytical confirmation.
How is AI used without replacing physics?
AI and machine-learning models can accelerate candidate generation, relaxation, and triage. High-impact candidates can then be escalated to periodic quantum and free-energy methods appropriate to the decision.
Does a low lattice energy guarantee a form will crystallize?
No. Thermodynamic competitiveness does not establish kinetic accessibility. We separate stability evidence from nucleation and process-access hypotheses and recommend targeted experimental testing.
Can existing PXRD or thermal data be incorporated?
Yes. Observed patterns, transition temperatures, desolvation events, and process history can constrain structural assignments, challenge predictions, and reprioritize experiments.
What defines a successful deliverable?
A useful project ends with a ranked risk narrative, traceable candidate structures, explicit limitations, and an experiment or control plan tied to the next program decision.
Plan your polymorph screen
Request a Project PlanStart Your Polymorph Project
Share your API structure, known solid-state evidence, and the development decision at risk. CD ComputaBio will propose a fit-for-purpose computational and experimental strategy.
Scientific References
- Neumann, M. A., van de Streek, J., Fabbiani, F. P. A., et al. Combined crystal structure prediction and high-pressure crystallization in rational pharmaceutical polymorph screening. Nature Communications 6, 7793 (2015). https://doi.org/10.1038/ncomms8793
- Zhou, D., Bier, I., Santra, B., et al. A robust crystal structure prediction method to support small molecule drug development with large scale validation and blind study. Nature Communications 16, 2210 (2025). https://doi.org/10.1038/s41467-025-57479-1
