A novel approach for calibrating quantum-chemical properties determined aspart of a high-throughput virtual screen to experimental analogs is presented.Information on the molecular graph is extracted through the use of extendedconnectivity fingerprints, and exploited using a Gaussian process to calibrateboth electronic properties such as frontier orbital energies, and optical gapsand device properties such as short circuit current density, open circuitvoltage and power conversion efficiency. The Bayesian nature of this processaffords a value for uncertainty in addition to each calibrated value. Thisallows the researcher to gain intuition about the model as well as the abilityto respect its bounds.
展开▼