Rectified Classifier Chains for Prediction of Antibiotic Resistance From Multi-Labelled Data With Missing Labels

Abstract

Predicting Antimicrobial Resistance (AMR) from genomic data has important implications for human and animal healthcare, and especially given its potential for more rapid diagnostics and informed treatment choices. With the recent advances in sequencing technologies, applying machine learning techniques for AMR prediction have indicated promising results. Despite this, there are shortcomings in the literature concerning methodologies suitable for multi-drug AMR prediction and especially where samples with missing labels exist. To address this shortcoming, we introduce a Rectified Classifier Chain (RCC) method for predicting multi-drug resistance. This RCC method was tested using annotated features of genomics sequences and compared with similar multi-label classification methodologies. We found that applying the eXtreme Gradient Boosting (XGBoost) base model to our RCC model outperformed the second-best model, XGBoost based binary relevance model, by 3.3% in Hamming accuracy and 7.8% in F1-score. Additionally, we note that in the literature machine learning models applied to AMR prediction typically are unsuitable for identifying biomarkers informative of their decisions; in this study, we show that biomarkers contributing to AMR prediction can also be identified using the proposed RCC method. We expect this can facilitate genome annotation and pave the path towards identifying new biomarkers indicative of AMR.

Explainable Deep Learning Approach for Multilabel Classification of Antimicrobial Resistance With Missing Labels

“…As part of gene identification, existing annotated genome sequences available from public databases 1,2 are used to annotate genes in a new sequence [9] with the help of tools such as the Basic Local Alignment Search Tool (BLAST) [10]. Once genes are extracted from a genomic sequence, this data is treated as input features for a machine learning approach to enable phenotypic prediction [6]; Several studies have been undertaken using machine learning algorithms to predict AMR from annotated genes [8], [11], [12], [13]. More genomics data is now available with the advancement in next-generation genomic sequencing.…”

“…Comparatively, few studies have been published on applying multilabel classification methods to genomics data for predicting multiple types of AMR [8], [16], [17], [18, p.], [19], [20], [21]. DeepGo [21] and DeepGoplus [22] are two popular methods applied to predicting multilabel protein classes from genomic sequences using deep learning methodologies; however, they do not predict AMR specifically nor handle missing labels as part of their operation.…”

“…Despite existing studies on handling missing labels on multilabel prediction, missing label scenarios for multidrug AMRs are considered in only a few research studies [8]. In this study, Rectified Classifier Chain (RCC) method for predicting multidrug resistance was proposed by internally modifying the existing classifier chain approach to handle missing labels.…”

“…This approach is faster than the conventional lab-based approach [4] and facilitates a personalised treatment plan, thus avoiding unnecessary antibiotic use that would otherwise exert selective pressure for resistance emergence. The development of Next-Generation Sequencing (NGS) technologies and increasing Graphical Processing Unit (GPU) accelerated data processing capability have enabled rapid and more cost-friendly prediction of AMR with machine learning [5], [6], [7], [8].…”

“…They need to be manually labelled with complex and time-consuming microbiological experiments. As the genomes were sequenced at different times and laboratory-based experiments were done only for the resistance of selected antibiotics, the resistances of other antibiotics are unknown and treated as missing labels [8]. This can result in poor performance as current deep learning models depend on correctly labelled data.…”