About ChiralCall

ChiralCall is a computational chiral enantiomer prediction tool built by a researcher with a background spanning computational chemistry, software development, and drug discovery research. The underlying prediction methodology was developed and validated over an extended independent research program.

The current production system covers 50 compound classes — including NSAIDs, opioids, β-blockers, CCBs, antifungals, antipsychotics, anticonvulsants, antimalarials, local anesthetics, herbicides, amino-alcohols, kinase inhibitors, pyrethroids, and axial-chirality atropisomers — verified against 1,051 compounds with published active enantiomer data at 98.5% accuracy (Wilson 95% CI: 97.7%–99.0%).

The methodology is proprietary. Predictions are derived from a first-principles computational approach — not a machine learning model trained on known chirality data. The validation was prospective and blind: compounds were selected and locked before any predictions were run.

A full methodology paper is in preparation.

ChiralCall was built by Matt Rusin, a technology entrepreneur who identified a gap in the pharmaceutical and agrochemical toolchain: no reliable way to predict which enantiomer of a chiral compound would be pharmacologically favored before synthesis. He developed the underlying prediction methodology through an extended independent research program, then built ChiralCall to make it accessible to researchers and teams worldwide.

ChiralCall is operated by Arroway Sciences. For inquiries, NDA requests, or access questions, use the contact form or reach out directly.

All accuracy statistics reflect honest out-of-sample testing. The original 70/72 prospective blind benchmark locked compounds before any predictions. The ChEMBL 36 cross-validation uses leave-one-out auto-direction learning where each compound's prediction excludes itself from the family direction vote. We are transparent about which sub-families are fully validated (Tier 1: Wilson CI lower >80% on N≥10) and which are exploratory (Tier 2: 100% exact on N≥6).

Predictions are intended for research and drug discovery use. They should be verified experimentally before any clinical, regulatory, or commercial decision.

ChiralCall is stateless by default. SMILES inputs are processed in memory and discarded immediately — nothing is stored unless you explicitly choose the Store or Contribute mode. No compound structures, prediction results, or user inputs are logged, cached, or retained on our servers in stateless mode.

The platform is hosted on Vercel with TLS encryption in transit. Authentication uses industry-standard OAuth and session tokens. API keys are scoped per user and can be revoked at any time from your dashboard.