Compounds Activity Prediction in Large Imbalanced Datasets with Substructural Relations Fingerprint and EEM

Abstract

Abstract-Modern drug design procedures involve the process of virtual screening, a highly efficient filtering step used for maximizing the efficiency of the preselection of compounds which are valuable drug candidates. Recent advances in introduction of machine learning models to this process can lead to significant increase in the overall quality of the drug designing pipeline.Unfortunately, for many proteins it is still extremely hard to come up with a valid statistical model. It is a consequence of huge classes disproportion (even 1000:1), large datasets (over 100,000 of samples) and restricted data representation (mostly high-dimensional, sparse, binary vectors). In this paper, we try to tackle this problem through three important innovations. First we represent compounds with 2-dimensional, graph representation. Second, we show how one can provide extremely fast method for measuring similarity of such data. Finally, we use the Extreme Entropy Machine which shows increase in balanced accuracy over Extreme Learning Machines, Support Vector Machines, one-class Support Vector Machines as well as Random Forest.Proposed pipeline brings significantly better results than all considered alternative, state-of-the-art approaches. We introduce some important novel elements and show why they lead to better model. Despite this, it should still be considered as a proof of concept and further investigations in the field are needed.

Learning from imbalanced data: open challenges and future directions

“…However, there is a lack of studies on the classification of extremely imbalanced datasets. In real-life applications such as fraud detection [61] or cheminformatics [12] we may deal with problems with imbalance ratio ranging from 1:1000 up to 1:5000. This poses new challenges to data preprocessing and classification algorithms, as they must be adjusted to such extreme scenarios.…”

Structure–activity relationship-based chemical classification of highly imbalanced Tox21 datasets

“…Structure–activity relationship (SAR) has been frequently used to predict the biological activities of chemicals from their molecular structures. One of the major challenges in SAR-based chemical classification or drug discovery is the extreme imbalance between active and inactive chemicals [ 1 ]. Despite the existence of as many as 10 7 commercially available molecules [ 2 ], there is almost always a skew in the distribution of molecules across the bioactivity landscape or toxicity classes.…”

“…This leads to low sensitivity and precision for the minority class [ 6 ], even though the minority class is usually of greater importance than the majority class [ 7 , 8 ]. In fields such as toxicology and disease diagnosis, bias towards the majority class may result in a higher rate of false negative predictions [ 1 ].…”

“…The present study was motivated by the scarcity of reported efforts in the application of the above-mentioned methods to the SAR-based chemical classification domain. We conducted a literature survey which only identified a few studies in this domain where cost-sensitive learning [ 28 , 29 ], resampling [ 29 , 30 ], conformal prediction [ 18 ] and extreme entropy machines [ 1 , 31 ] were employed to specifically deal with data imbalance. Although predictive modeling was improved for certain datasets, a consistent performance enhancement was not observed as a result of resampling and algorithm modification.…”

Potential Anchoring for imbalanced data classification

“…The problem of data imbalance is ubiquitous in practical applications, affecting domains such as cancer malignancy grading [8], industrial systems monitoring [9], fraud detection [10], behavioral analysis [11] and cheminformatics [12]. Furthermore, data imbalance typically leads to the more costly type of error, for instance by inducing false negatives in the medical problem domain.…”