In collaboration with a major pharmaceutical company, we at Atomwise apply AtomNet, a deep convolutional neural network for bioactivity prediction, to quantitatively predict small molecule binding affinity over the lifetime of various projects spanning 9 diverse protein targets. We evaluate the impact AtomNet would have if we were to enter the project at 4 different time points, spanning from blind predictions before the project starts, to using two-thirds of the project series data to predict the final third. To conclude we answer the question: what is the effect of more client data on our blinded prediction performance?

Evaluating lead optimization performance of a structure-based convolutional neural network

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.