Turbo prediction: a new approach for bioactivity prediction

Nowadays, activity prediction is key to understanding the mechanism-of-action of active structures discovered from phenotypic screening or found in natural products. Machine learning is currently one of the most important and rapidly evolving topics in computer-aided drug discovery to identify and design new drugs with superior biological activities. The performance of a predictive machine learning model can be enhanced through the optimal selection of learning data, algorithm, algorithm parameters, and ensemble methods. In this article, we focus on how to enhance the prediction model using the learning data. However, get an option to add more and accurate data is not easy and available in many cases. This motivated us to propose the turbo prediction model, in which nearest neighbour structures are used to increase prediction accuracy. Five datasets, well known in the literature, were used in this article and experimental results show that turbo prediction can improve the quality prediction of the conventional prediction models, particularly for heterogeneous datasets, without any additional effort on the part of the user carrying out the prediction process, and at a minimal computational cost.