Changelog
What we shipped and when. No vaporware — everything here is live in production.
May 2026
Cost calculator for chiral resolution savings
Interactive ROI calculator on the homepage — input your quarterly compound volume and per-resolution cost, see how much ChiralCall saves by prioritizing the correct enantiomer first.
Decision memos and shareable predictions
Copy a formatted decision summary from any prediction — paste it into Slack, email, or your lab notebook. Batch comparison reports now include a one-click "Copy Decision Memo" for internal distribution.
Blinded benchmark comparison tool
CRO teams can now compare ChiralCall predictions against their own ground truth, entirely client-side. Your answers never leave your browser.
Safer predictions for symmetric scaffolds
DHP-C4-symmetric compound class split — dihydropyridines with C4 symmetry now return a separate, properly validated prediction instead of being silently merged with asymmetric DHPs.
Direction fix for diaryl-cyclic-amine-AH class
Corrected the predicted active enantiomer direction for the diaryl-cyclic-amine-AH compound class after identifying an inversion error during expanded validation.
Leave-one-out cross-validation across all classes
Every compound class now validated with LOO cross-validation on ChEMBL 36 data. Overall served accuracy: 98% across 691 ChEMBL compounds.
April 2026
Works with any SMILES notation
Switched from brittle text-substring matching to RDKit SMARTS substructure matching. Your SMILES string can use any valid notation — Kekulé, aromatic, any atom ordering — and ChiralCall resolves it correctly.
Tells you when it's uncertain
Specificity scoring automatically detects when a compound matches multiple compound classes. Instead of silently picking one, ChiralCall flags the ambiguity and selects the most structurally specific match.
Meets regulatory documentation requirements
API responses now include per-class validation statistics — accuracy, Wilson 95% confidence intervals, LOO results, and compound class size — so predictions can be cited in regulatory filings without manual lookup.
Calibrated Confidence Score (CCS) in API
Every prediction now includes a Calibrated Confidence Score (0–100) computed via logistic regression over structural features. CCS tiers (very high, high, medium, low) give you a per-compound reliability estimate.
CRO Pilot program launched
Free 60-day pilot for CRO teams: submit up to 150 compounds, get predictions with full validation data, compare against your own ground truth. No credit card, no commitment.
March 2026
3D molecule viewer on prediction pages
Interactive 3D visualization of predicted stereochemistry on every shareable prediction page. Real molecular structures from PubChem with drag-to-rotate and pinch-to-zoom.
Wrong predictions browser
Full transparency: browse every wrong prediction ChiralCall has ever made, filterable by compound class, with per-class accuracy bars and error breakdowns.
February 2026
Dashboard and API v1
User dashboard for managing predictions and API keys. RESTful API v1 with batch processing, authentication, rate limiting, and programmatic access to all prediction data.
Kinase atropisomer support
Dedicated compound class for atropisomeric kinase inhibitors — a scaffold type that most chirality tools ignore entirely.
January 2026
Agrochemical compound classes
13 agrochemical compound classes including pyrethroids, triazole fungicides, and organophosphates. ChiralCall now covers both pharmaceutical and agricultural chirality.
Coming Soon
Python package
pip install chiralcall — full API bindings, batch processing, and notebook integration.
R package
install.packages("chiralcall") — tidyverse-compatible interface for R users.
Custom compound class development
Paid service for proprietary scaffolds not in our public library. We validate on your data, you get a private compound class.
More compound classes every month
Currently at 53 compound classes and growing. Community-requested scaffolds are prioritized — request yours below.
Don't see your compound class?
Request a new scaffold and we'll prioritize validation if it has ≥10 known compounds.