At low doses of the sensitisers, the sensitisation may be greater in lines expressing low levels of P-glycoprotein than in lines showing high levels.
Although these results are very promising, it is currently unpredictable whether the effect of these identified sensitisers is selective to tumour cells or if they would also sensitise non-transformed cells to TRAIL and thus cause toxicity.
False positive (FP) = number of non sensitisers predicted as sensitisers.
False negative (FN) = number of sensitisers predicted as non sensitisers.
This gave a dataset with 108 non-sensitisers and 101 sensitisers, (moderate/strong/extreme categories; EC3 values >10).
True negative (TN) = number of non sensitisers correctly predicted.
LLNA Class Total compounds Training Test Non-sensitisers NC 42 34 8 Sensitisers Weak+Moderate+Strong+Extreme 167 133 34 Non-sensitisers Weak+NC 108 87 21 Sensitisers Moderate+Strong+Extreme 101 80 2101 209 101 42.
Adaptive Fuzzy Partition (AFP) modelling used a binary split of data, (sensitisers and non-sensitisers) in which only those compounds classed as "NC" in the original data were deemed to be non-sensitisers.
MLP modeling was applied to the binary class data in which both non-sensitisers and weak sensitisers were classified as inactive (Table 6).
For the multi-layer perceptron (MLP) modelling, non-sensitisers also included those compounds that were originally classified as weak sensitisers, (Table 6).
The technique described can be used for testing the suitability of dyes as sensitisers.
