Monte Carlo-based SMILES-driven QSPR modeling of Organic Chromophores across the visible-to- NIR spectral range

Developing organic chromophores with tailored optical properties is a key objective in material chemistry. In the present study, a dataset of 130 organic chromophores was used to build QSPR models for absorption and emission maxima using CORAL. The models were generated using SMILES derived descriptors optimised through Monte-Carlo method. The model stability was evaluated using non-identical data splits and multiple target functions IIC, CII, CCCP and correlation balance. The resulting models showed good predictive performance based on internal and external validation parameters such as R², Q², MAE and standard error. Among the developed models, A18 generated using the CCCP-based target function (TF-4) is the best model for absorption maxima with validation statistics of R² = 0.8404, Q² = 0.8261, MAE = 39.4, AvR_m² = 0.7703 and ΔR_m² = 0.1077. For emission maxima, F10 provided the best overall performance, giving R² = 0.8332, Q = 0.8192, MAE = 32.4, AvR_m² = 0.7615 and ΔR_m² = 0.0019. These results confirm the predictive reliability of the SMILES-based CORAL models. Applicability domain assessment indicated that most of the compounds were within the reliable prediction range of the models. These results suggest that the proposed models can be used as a practical tool for estimating chromophores spectral properties and guiding the design of near-infrared materials.