Optimal ECG-lead selection increases generalizability of deep learning on ECG abnormality classification

Deep learning (DL) has achieved promising performance in detecting common abnormalities from the 12-lead electrocardiogram (ECG). However, diagnostic redundancy exists in the 12-lead ECG, which could impose a systematic overfitting on DL, causing poor generalization. We, therefore, hypothesized that finding an optimal lead subset of the 12-lead ECG to eliminate the redundancy would help improve the generalizability of DL-based models. In this study, we developed and evaluated a DL-based model that has a feature extraction stage, an ECG-lead subset selection stage and a decision-making stage to automatically interpret multiple common ECG abnormality types. The data analysed in this study consisted of 6877 12-lead ECG recordings from CPSC 2018 (labelled as normal rhythm or eight types of ECG abnormalities, split into training (approx. 80%), validation (approx. 10%) and test (approx. 10%) sets) and 3998 12-lead ECG recordings from PhysioNet/CinC 2020 (labelled as normal rhythm or four types of ECG abnormalities, used as external text set). The ECG-lead subset selection module was introduced within the proposed model to efficiently constrain model complexity. It detected an optimal 4-lead ECG subset consisting of leads II, aVR, V1 and V4. The proposed model using the optimal 4-lead subset significantly outperformed the model using the complete 12-lead ECG on the validation set and on the external test dataset. The results demonstrated that our proposed model successfully identified an optimal subset of 12-lead ECG; the resulting 4-lead ECG subset improves the generalizability of the DL model in ECG abnormality interpretation. This study provides an outlook on what channels are necessary to keep and which ones may be ignored when considering an automated detection system for cardiac ECG abnormalities. This article is part of the theme issue ‘Advanced computation in cardiovascular physiology: new challenges and opportunities’.

Electrocardiogram lead selection for intelligent screening of patients with systolic heart failure

Electrocardiogram (ECG)-based intelligent screening for systolic heart failure (HF) is an emerging method that could become a low-cost and rapid screening tool for early diagnosis of the disease before the comprehensive echocardiographic procedure. We collected 12-lead ECG signals from 900 systolic HF patients (ejection fraction, EF < 50%) and 900 individuals with normal EF in the absence of HF symptoms. The 12-lead ECG signals were converted by continuous wavelet transform (CWT) to 2D spectra and classified using a 2D convolutional neural network (CNN). The 2D CWT spectra of 12-lead ECG signals were trained separately in 12 identical 2D-CNN models. The 12-lead classification results of the 2D-CNN model revealed that Lead V6 had the highest accuracy (0.93), sensitivity (0.97), specificity (0.89), and f1 scores (0.94) in the testing dataset. We designed four comprehensive scoring methods to integrate the 12-lead classification results into a key diagnostic index. The highest quality result among these four methods was obtained when Leads V5 and V6 of the 12-lead ECG signals were combined. Our new 12-lead ECG signal–based intelligent screening method using straightforward combination of ECG leads provides a fast and accurate approach for pre-screening for systolic HF.

“Branched Tail” Oxyquinoline Inhibitors of HIF Prolyl Hydroxylase: Early Evaluation of Toxicity and Metabolism Using Liver-on-a-chip

Background: “Branched tail” oxyquinolines, and adaptaquin in particular, are potent HIF prolyl hydroxylase inhibitors showing promising results in in vivo hemorrhagic stroke models. The further improvement of the potency resulted in identification of a number of adaptaquin analogs. Early evaluation of toxicity and metabolism is desired right at the step of lead selection. Objective: The aim of the study is to characterize the toxicity and metabolism of adaptaquin and its new improved analogs. Method: Liver-on-a-chip technology with differentiated HepaRG cells followed by LC-MS detection of the studied compounds and metabolites of the P450 substrate-inhibitor panel for CYP2B6, CYP2C9, CYP2C19, and CYP3A4. Results: The optimized adaptaquin analogs show no toxicity up to a 100-fold increased range over EC50. The drugs are metabolized by CYP3A4 and CYP2B6 as shown with the use of the cytochrome P450 substrate-inhibitor panel designed and optimized for preclinical evaluation of drugs’ in vitro biotransformation on a 3D human histotypical cell model using “liver-on-a-chip” technology. Activation of CYP2B6 with the drugs tested has been observed. A scheme for adaptaquin oxidative conversion is proposed. Conclusion: The optimized adaptaquin analogs are suitable for further preclinical trials. Activation of CYP2B6 with adaptaquin and its variants points to a potential increase in Tylenol toxicity if administered together.

New computational studies to support cyclin-dependent kinase 9 inhibitor screening and design

Aims: Cyclin-dependent kinase 9 (CDK9) plays a major role in the regulation of transcription. Its overexpression -which occurs in several types of cancer- increases the levels of certain antiapoptotic proteins that can lead to tumorigenesis, therefore the identification of new, more potent and more selective inhibitors is essential. Methods: In this study we present a computational approach, which can facilitate lead selection and optimization. Results: First, a pharmacophore hypothesis based on the active compounds has been developed to identify the key features for the ligand-target interaction. This was followed by the docking of the compounds into the active site of CDK9, the poses and interactions with the amino acids were compared with those of the co-crystallized ligand. The mode of their binding further explained the characteristics of these inhibitors while the docking scores can be a factor in the selection of active compounds in the future. Finally, a field-based QSAR model was also created, to predict the activity of inhibitor candidates. Conclusion: With our current work we deepened our knowledge about the interactions between CDK9 and its inhibitors, which can contribute to the discovery of novel CDK9 inhibitors.

Lead Optimization in the Drug Discovery Process

Discovering a new drug molecule against disease is the main objective of drug discovery. Lead optimization is one of the important steps and acts as a starting point. Over the years, it has significantly changed the drug discovery process. Its main focus is the development of preclinical candidates from “Hit” or “Lead.” Lead optimization comprises lead selection and optimization, drug candidate confirmation, and preclinical drug characterization. Lead optimization process can improve the effectiveness towards its target potency, selectivity, protein binding, pharmacokinetic parameters, and to develop a good preclinical candidate. Lead optimization from high-throughput screening to identification of clinical drug candidate is a seamless process that draws new techniques for accelerated synthesis, purification, screening from iterative compound libraries, validation, and to deliver clinical drug candidate with limited human resources. In conclusion, lead optimization phase is done under the suggestion that the optimized lead molecule will have activity against a particular disease.