Background: Cardiac channelopathies such as catecholaminergic polymorphic tachycardia and long QT syndrome predispose patients to fatal arrhythmias and sudden cardiac death. As genetic testing has become common in clinical practice, variants of uncertain significance (VUS) in genes associated with catecholaminergic polymorphic ventricular tachycardia and long QT syndrome are frequently found. The objective of this study was to predict pathogenicity of catecholaminergic polymorphic ventricular tachycardia-associated RYR2 VUS and long QT syndrome-associated VUS in KCNQ1 , KCNH2 , and SCN5A by developing gene-specific machine learning models and assessing them using cross-validation, cellular electrophysiological data, and clinical correlation. Methods: The GENe-specific EnSemble grId Search framework was developed to identify high-performing machine learning models for RYR2 , KCNQ1 , KCNH2 , and SCN5A using variant- and protein-specific inputs. Final models were applied to datasets of VUS identified from ClinVar and exome sequencing. Whole cell patch clamp and clinical correlation of selected VUS was performed. Results: The GENe-specific EnSemble grId Search models outperformed alternative methods, with area under the receiver operating characteristics up to 0.87, average precisions up to 0.83, and calibration slopes as close to 1.0 (perfect) as 1.04. Blinded voltage-clamp analysis of HEK293T cells expressing 2 predicted pathogenic variants in KCNQ1 each revealed an ≈80% reduction of peak Kv7.1 current compared with WT. Normal Kv7.1 function was observed in KCNQ1-V241I HEK cells as predicted. Though predicted benign, loss of Kv7.1 function was observed for KCNQ1-V106D HEK cells. Clinical correlation of 9/10 variants supported model predictions. Conclusions: Gene-specific machine learning models may have a role in post-genetic testing diagnostic analyses by providing high performance prediction of variant pathogenicity.
Explanation methods facilitate the development of models that learn meaningful concepts and avoid exploiting spurious correlations. We illustrate a previously unrecognized limitation of the popular neural network explanation method Grad-CAM: as a side effect of the gradient averaging step, Grad-CAM sometimes highlights locations the model did not actually use. To solve this problem, we propose HiResCAM, a novel class-specific explanation method that is guaranteed to highlight only the locations the model used to make each prediction. We prove that HiResCAM is a generalization of CAM and explore the relationships between HiResCAM and other gradient-based explanation methods. Experiments on PASCAL VOC 2012, including crowd-sourced evaluations, illustrate that while HiResCAM's explanations faithfully reflect the model, Grad-CAM often expands the attention to create bigger and smoother visualizations. Overall, this work advances convolutional neural network explanation approaches and may aid in the development of trustworthy models for sensitive applications.
12 August 2022 in Artificial Intelligence in Medicine
Although board games and video games have been studied for decades in artificial intelligence research, challenging word games remain relatively unexplored. Word games are not as constrained as games like chess or poker. Instead, word game strategy is defined by the players' understanding of the way words relate to each other. The word game Codenames provides a unique opportunity to investigate common sense understanding of relationships between words, an important open challenge. We propose an algorithm that can generate Codenames clues from the language graph BabelNet or from any of several embedding methods - word2vec, GloVe, fastText or BERT. We introduce a new scoring function that measures the quality of clues, and we propose a weighting term called DETECT that incorporates dictionary-based word representations and document frequency to improve clue selection. We develop BabelNet-Word Selection Framework (BabelNet-WSF) to improve BabelNet clue quality and overcome the computational barriers that previously prevented leveraging language graphs for Codenames. Extensive experiments with human evaluators demonstrate that our proposed innovations yield state-of-the-art performance, with up to 102.8% improvement in precision@2 in some cases. Overall, this work advances the formal study of word games and approaches for common sense language understanding.