Single-molecule research techniques such as patch-clamp electrophysiology deliver unique biological insight by capturing the movement of individual proteins in real time, unobscured by whole-cell ensemble averaging. The critical first step in analysis is event detection, so called “idealisation”, where noisy raw data are turned into discrete records of protein movement. To date there have been practical limitations in patch-clamp data idealisation; high quality idealisation is typically laborious and becomes infeasible and subjective with complex biological data containing many distinct native single-ion channel proteins gating simultaneously. Here, we show a deep learning model based on convolutional neural networks and long short-term memory architecture can automatically idealise complex single molecule activity more accurately and faster than traditional methods. There are no parameters to set; baseline, channel amplitude or numbers of channels for example. We believe this approach could revolutionise the unsupervised automatic detection of single-molecule transition events in the future.
There remains a need for tools leveraging widely accessible data compatible with mobile health platforms to provide simple and broadly comprehensible cardiovascular risk estimations. We streamlined the previously published DiCAVA model, originally developed in the UK Biobank (UKB), into a 30-feature tool (DiCAVA-Lite) to predict incident myocardial infarction, stroke, heart failure, angina, and transient ischemic attack in primary prevention settings. We then validated DiCAVA-Lite in the All of Us (AoU) using Cox proportional hazard regression models and an incident composite cardiovascular outcome. We then translate the DivCAVA-Lite score to heart age in both UKB and AoU, which communicates the age of a healthy person with equivalent cardiovascular risk to that of the index individual. In the UKB dataset (n = 466,052), the mean age was 56 years (SD: 8) with 55.9% females, while in AoU (n = 216,985), the mean age was 50 years (SD: 17) with 63.2% females. The DiCAVA-Lite model demonstrated moderate discrimination, with the area under the curve (AUC) of 0.740 (95% CI: 0.735-0.745) (vs. 0.747 [95% CI: 0.744-0.748] for DiCAVA) in UKB and 0.770 (95% CI: 0.765-0.773) in external validation in AoU. Heart age demonstrated discriminative performance equivalent to the DiCAVA-Lite model. In UKB, the mean heart age was 60 among individuals who developed events vs. 55 in those who did not. In AoU, the corresponding ages were 57 and 50, respectively. These findings support DiCAVA-Lite and the corresponding heart age as an accessible and interpretable tool for CVD risk assessment and communication in the US and UK primary prevention populations.
Single-molecule research such as patch-clamp recording delivers unique biological insight by capturing the movement of individual proteins in real time, unobscured by whole-cell ensemble averaging. The critical first step in analysis is event detection, so called “idealisation”, where noisy raw data are turned into discrete records of protein movement. To date there have been practical limitations in patch-clamp data idealisation; high quality idealisation is typically laborious and becomes infeasible and subjective with complex biological data containing many distinct native single ion channel proteins gating simultaneously. Here we show a deep learning model based on convolutional neural networks and long short-term memory architecture can automatically idealise complex single molecule activity more accurately and faster than traditional methods. There are no parameters to set; baseline, channel amplitude or numbers of channels for example. We believe this approach could revolutionise the unsupervised automatic detection of single-molecule transition events in the future.