MolScore-Antibiotics discriminates between antibiotics and non-antibiotics. The MolScore-Antibiotics of a compound measures the probability of having antibiotic activity and is defined as a value between 0 and 1.
MolScores-Antibiotics is able to detect and prioritise candidates for clinical development. The expert system has been validated with novel antibacterials which are now in clinical trials.
MolScore-Antibiotics near 0: => lowest predicted antibiotic activity
MolScore-Antibiotics near 1: => highest predicted antibiotic activity
The expert system uses a number of different strategies to find novel antibiotics. Models, based on neural networks, pharmacophore models, structure-ADME-relations, decision-trees and affinity prediction to antibiotic drug targets have been included, see science & technology. It is complicated to calculate but easy to use!
Our methods have been proven to predict the antibiotic activity with an extremely low error. The error rate, estimated with an independent validation data set, was less than 5 %. This shows the excellent generalisation ability of the models used.
We have analysed whether MolScore-Antibiotics could select antibiotic drugs from the blockbusters of the last five years. All antibiotic blockbusters had MolScore-Antibiotics results higher than 0.996, see proof of principle for details. MolScore-Antibiotics has also detected future drugs, which are now in clinical trials.
MolScore-Antibiotics can be used to guide the buying or selecting of compounds for focussed biological screening (prioritisation of compounds from large compound collections). Our expert system revealed that many compound databases from external suppliers only have a small amount of compounds with antibiotic activity. Interesting compounds can be easily cherry-picked, see our example (selection of antibiotics from an external database of 195.064 compounds).
Expert system for antibacterial research
==> Hit detection & validation
==> Lead selection & prioritisation