Tuning gradient boosting for imbalanced bioassay modelling with custom loss functions

Abstract

While in the last years there has been a dramatic increase in the number of available bioassay datasets, many of them suffer from extremely imbalanced distribution between active and inactive compounds. Thus, there is an urgent need for novel approaches to tackle class imbalance in drug discovery. Inspired by recent advances in computer vision, we investigated a panel of alternative loss functions for imbalanced classification in the context of Gradient Boosting and benchmarked them on six datasets from public and proprietary sources, for a total of 42 tasks and 2 million compounds. Our findings show that with these modifications, we achieve statistically significant improvements over the conventional cross-entropy loss function on five out of six datasets. Furthermore, by employing these bespoke loss functions we are able to push Gradient Boosting to match or outperform a wide variety of previously reported classifiers and neural networks. We also investigate the impact of changing the loss function on training time and find that it increases convergence speed up to 8 times faster. As such, these results show that tuning the loss function for Gradient Boosting is a straightforward and computationally efficient method to achieve state-of-the-art performance on imbalanced bioassay datasets without compromising on interpretability and scalability. Graphical Abstract

Biomedical data analyses facilitated by open cheminformatics workflows

“…This process has been encouraged for research papers submitted to J. Cheminf. as demonstrated by these examples [ 22 , 23 ].…”

Affine combination‐based over‐sampling for imbalanced regression

“…23 As a result, the imbalance of chemical datasets can present challenges for modeling and analysis, and it is important to use suitable techniques and methods to handle imbalanced data. 17,18,22,[24][25][26][27][28] Research in the field of imbalanced learning has provided better data analysis techniques to address imbalanced problems in these areas. In the imbalanced regression prediction modeling scenario, there are two characteristics 3 : (i) the skewed distribution of continuous response variables and (ii) a domain preference for underrepresented instances.…”

“…The majority of in silico models that predict toxicity profiles and bioactivities were dependent on imbalanced datasets 23 . As a result, the imbalance of chemical datasets can present challenges for modeling and analysis, and it is important to use suitable techniques and methods to handle imbalanced data 17,18,22,24–28 . Research in the field of imbalanced learning has provided better data analysis techniques to address imbalanced problems in these areas.…”

ASER: Adapted squared error relevance for rare cases prediction in imbalanced regression

“…Therefore, there is an urgent need for new methods to address this problem of imbalance. [13][14][15][16] For example, cancer lectins are those that firmly recognize specific types of proteins, which initiate the tolerance, development, metastasis, and spread of cancer cells. We need to separate cancer lectins and noncancer lectins to better study cancer treatment.…”

“…In the past few years, the number of available bioassay datasets has sharply increased, but many of them have extremely imbalanced distributions between active and inactive compounds or imbalance of sample number between the new and the previous experimental data. Therefore, there is an urgent need for new methods to address this problem of imbalance 13–16 . For example, cancer lectins are those that firmly recognize specific types of proteins, which initiate the tolerance, development, metastasis, and spread of cancer cells.…”