Попов Никита Сергеевич
One-Shot Learning in Bioinformatics
Прикладная математика и информатика
This paper discusses one of the possible problems occurring in the field of drug discovery–selectivity maximization, which can be explained as increasing desired effects while minimizing potential side effects. However, while the definition of this problem might seem simple, in order to solve it one has to overcome two main obstacles: measuring the selectivity index and optimizing over molecular space. This paper proposes ways to tackle both of these complications and combine the parts into an optimization pipeline. Additionally, this paper touches on the neighboring topics, such as using one-shot learning to train the model and briefly discusses potential concerns with optimizing towards an approximation.