andykoswara / adme_tox Goto Github PK
View Code? Open in Web Editor NEWPython script that implements a random forest algorithm to predict several ADME-Tox classifications of bioactive molecules accompanied with a visualization technique called Uniform Manifold Approximation Projection (UMAP) [1]. This work is an amalgamation of previous great work by fellow researchers [2-5] with an extension towards our own research on predicting molecular ion fragmentation by a mass spectrometer (MS). In particular, we investigated the impact of different molecular encodings on the algorithm's prediction accuracy, sensitivity and specificity.
License: BSD 3-Clause "New" or "Revised" License