Advanced QSRR modeling of peptides behavior in RPLC.

摘要: In QSRR the retention is modeled as a function of structural or molecular descriptors. Since datasets can be very large selection informative variables often required. But beside question which subset (descriptors) produces optimum predictions one should answer question: good prediction used in community even if physical meaning applied descriptors hard to interpret? The main focus this paper put on different modeling methodologies and approaches. Besides widely multiple linear regression (MLR), these include partial least squares (PLS), uninformative variable elimination (UVE-PLS), genetic algorithms (GA) prior MLR PLS. comparison will predictive performance but also found most important for chromatographic peptides. results study showed that stepwise-MLR UVE-PLS are producing better than rest studied methodologies. From selected by various see information mechanism RPLC was given 2D-, 3D-descriptors from empirical equations, bring about hydrogen-bonding properties, size, complexity. Overall, considered data set models were predicting peptides best.