Assessment of taxol activity: Bioavailability in human physiological fluids and QSAR of taxol analogues based on a neural network design

Abstract

Taxol is an anticancer agent that is highly protein bound. This feature is addressed in the first chapter by developing a competitive, enzyme-linked immunosorbent assay (ELISA) to quantify bioavailability of taxol in human fluids: plasma, plasma ultrafiltrate and salivary fluids. An optimization procedure decreased the detection limit to pg/ml. The working range encompasses five orders in magnitude, i.e. from pg/ml to 100 ng/ml. ELISA was then applied to detect taxol in samples from a cancer patient. The results show the taxol concentration in plasma (sub-μg/ml), ultrafiltrate (ng/ml) and salivary fluids (sub-ng/ml).

The second chapter describes a new class of immunosensors in which the binding event is directly transduced, so there is no requirement for a chemical label. The ligand is Staphylococcal Enterotoxin B (SEB), while the receptor is the anti-SEB antibody. A high sensitivity radioimmunoassay was developed to detect about 1 pmol/cm2 of bioactive SEB-antibody, immobilized on the sensor surface. This technology has the potential for a low cost diagnostic assay and could be used in the medical and food industries.

Chapter three describes a back-propagation neural network (BPNN) design for 50 taxol analogues. Initial system consists of 27 calculated structural descriptors, while the outputs are the measured antitumor activities against 4 types of cancer (breast, ovarian, lung and the overall average GI50 ). The optimization process leads to dimensional reduction with increased accuracy (in comparison to other parametric or non-parametric techniques), in terms of correlation, i.e. 0.831 ovarian, 0.945 lung, 0.913 breast cancer, and 0.886 for the index G150. Calculated molecular properties and the anticancer activities are the only required input and output variables, so that the potential of the compound can be established prior to the synthesis.

Quantitative structure-activity relationships (QSAR) for a set of 61 taxol analogues previously excluded from a BPNN design are explained in chapter four. The input data are calculated descriptors (lipophilicity, molar refractivity, dipole moment plus dipole vectors, steric plus conformation energies, and heat of formation). High antitumor activity is predicted for taxol analogues with a substituent propionic, 1-methyl-2-pyrrolecarboxylic, or crotonic acid in position C-10.