Novel green sorbent Aerva Javanica for dyes remediation from aqueous media and future prediction through QSPR modeling

Pollutant removal from different sources, especially water pollutants, is a major concern in today's world. Various methodologies were being utilized to purify water resources because water scarcity is a major problem of the twenty-first century. In the current study, we searched and discovered novel green, renewable, cost-effective, environmentally benign, and readily available adsorbent Aerva Javanica (A. Javanica), for remediation of selected fourteen (14) dyes from aqueous media. A quantitative structure–property relationship model was developed to relate the structural properties of dyes with the percentage adsorption based on data and information which was collected by the adsorption study of these selected dyes. The structures of these dyes were optimized using MOPAC2016 and DRAGON software and further used for descriptors calculation. Initially, 1666 descriptors were calculated. The heuristic method was employed to select significant descriptors, and after pre-reducing steps, we left with 36 descriptors. Stepwise multiple linear regression (SMLR) analysis followed by artificial neural network (ANN) was employed on these descriptors for model generation. The SMLR-ANN model has shown better predictive ability (R² = 0.9907) than only the SMLR model (R² = 0.9778). Validation of the model was performed by internal validation by measuring cross-validation squared correlation coefficient (Q² = 0.921) and external validation (predictive R² = 0.9211). Significant descriptors calculated for modeling include Mor10e, HATS5p, nHAcc, E3m, and GATS5v. All these descriptors are built on basic molecular properties like molecular structure, electronegativity, and polarity. So the model is significant in this regard as it relates basic molecular properties with their adsorption potential. Generated models have shown good robustness, stability, and predictive ability.