Deep learning for low-data drug discovery: hurdles and opportunities

Abstract

Deep learning is becoming increasingly relevant in drug discovery, from de novo design to protein structure prediction and synthesis planning. However, it is often challenged by the small data regimes typical of certain drug discovery tasks. In such scenarios, deep learning approaches – which are notoriously ‘data-hungry’ – might fail to live up to their promise. Developing novel approaches to leverage the power of deep learning in low-data scenarios is sparking great attention, and future developments are expected to propel the field further. This minireview provides an overview of recent low-data-learning approaches in drug discovery, analyzing their hurdles and advantages. Finally, we venture to provide a forecast of future research directions in low-data learning for drug discovery.

Machine learning-guided high throughput nanoparticle design

“…Since the number of available data is oen highly limited, specic machine learning strategies like active machine learning are particularly suited for this task. [23][24][25][26][27] By operating in an iterative fashion, active machine learning uses model predictions to decide which samples should be screened and added to the training data to update the model in the next cycle. 28,29 This allows models to reach a desired response faster by screening fewer samples.…”

“…Finally, machine learning can be used to guide nanoparticle development 21,22 with the aim of reducing the number of nanoparticle formulations needed to optimize a response. Since the number of available data is often highly limited, specific machine learning strategies like active machine learning are particularly suited for this task [23][24][25][26][27] . By operating in an iterative fashion, active machine learning uses model predictions to decide which samples should be screened and added to the training data to update the model in the next cycle 28,29 .…”