COMPARISON OF SEVERAL REGRESSION METHODS APPLIED IN DISPERSE DYE-CELLULOSE BINDING

Quantitative structure-affinity relationships were applied to a series of 27 disperse dyes by partial least squares (PLS) analysis and compared to previously published MLR (multiple linear regression), MTD (minimum steric difference) and CoMFA (comparative molecular field analysis) results. Calculated 0D, 1D and 2D structural dye features were correlated to their affinity for cellulose by PLS. A robust model (R2X(cum) = 0.617, R2Y(cum) = 0.959, Q2(cum) = 0.953) with predictive power was obtained from these correlations. Better statistical results were achieved in the PLS model, in comparison to the previous MLR, MTD and CoMFA results, but the three-dimensional models obtained by CoMFA gave more information on the dye-cellulose specific interactions.