Clinical genetic testing for hereditary cardiovascular diseases is recommended by many cardiovascular groups (Musunuru et al., 2020; Wilde et al., 2022). Genetic test results can be important for patient medical management and for the care for family members (Cirino et al., 2017). Appropriate classification of genetic variants is a critical component of this process and ultimately impacts patient and family outcomes (Care et al., 2017; Phillips et al., 2005). The American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) created recommendations for the classification of pathogenicity of variants in genes associated with monogenic disease (Richards et al., 2015). These recommendations include defining the criteria for evidence used in classification as well as providing a framework for weighing and combining different types of evidence for the classification. Despite this standardized approach to interpretation, analysis and appropriate classification of variants remain challenging across disease contexts in clinical genetics (McInnes et al., 2021).

In this issue of Annals of Noninvasive Electrocardiology, Younis et al. (2023) report that computational genetic variant prediction tools could identify the majority of pathogenic variants in congenital long QT syndrome (LQTS) 1–3. The authors also found that the computational scores did not predict clinical outcomes.

While it is encouraging that the variant prediction tools correlated with pathogenicity in this study, it is also important to note that determination of variant pathogenicity includes multiple types of evidence, including variant prediction evidence. Importantly, computational in silico predictors alone should not be used to classify the pathogenicity of a variant, but can be used as one piece of evidence in the classification of a genetic variant. The ACMG/AMP recommendations specify that using computational predictors are “supporting” level of evidence for or against pathogenicity using criteria PP3 and BP4 (Care et al., 2017). Supporting-level evidence must be combined with other more substantial lines of evidence to classify the variant. Furthermore, a recent manuscript by Pejaver et al. (2022) provided evidence for redefining how computational tools can be used to provide evidence for or against pathogenicity of variants using the Bayesian adaptation of the ACMG/AMP framework. This work showed that the tools can provide stronger than supporting evidence and the computational tools varied in their ability to reach these levels of evidence. These authors also pointed out that it is important to select a single tool to use for PP3/BP4 missense evidence to avoid biases in results selection.

In addition, although the terms continue to be used interchangeably in the literature, recently, the clinical genetics and genomics community has begun to distinguish the differences between the terms variant classification and variant interpretation. Variant classification is defined as the process of evaluating pathogenicity of a variant, while variant interpretation refers to the process of clinical integration of the genetic test results with patient clinical characteristics and family history to arrive at a diagnosis (Biesecker et al., 2018). These nuances are challenging in the field and it will continue to be important to carefully define and use such terms clearly in clinical genomic work in the future.

Younis et al. (2023) also found that the computational in silico tools did not predict clinical outcomes and conclude that variant location/functional analysis are needed for more accurate risk interpretation. The potential use of computational in silico tools outside of the variant classification process, such as using these tools as indicators of clinical severity, is much less clear. Accurate assessment of risk interpretation in LQTS and other heritable cardiovascular diseases continues to be challenging, complicated by variable expressivity and incomplete penetrance (Lankaputhra & Voskoboinik, 2021). In addition to additional location/functional data that may help clarify risk stratification among individuals with pathogenic and likely pathogenic variants in associated genes, further population-based and genotype-first approaches may also be used to further assess these complex issues related to variant interpretation (Wilczewski et al., 2023).

许多心血管团体建议对遗传性心血管疾病进行临床基因检测（Musunuru 等人，  2020；Wilde 等人，  2022）。基因检测结果对于患者的医疗管理和家庭成员的护理非常重要（Cirino 等人，  2017）。遗传变异的适当分类是这一过程的关键组成部分，并最终影响患者和家庭的结果（Care 等人，  2017；Phillips 等人，  2005）。美国医学遗传学和基因组学学院以及分子病理学协会 (ACMG/AMP) 针对与单基因疾病相关的基因变异的致病性分类提出了建议（Richards 等，2015  ）。这些建议包括定义分类中使用的证据标准，以及提供权衡和组合不同类型证据进行分类的框架。尽管采用了这种标准化的解释方法，但在临床遗传学的疾病背景下，变异的分析和适当分类仍然具有挑战性（McInnes et al.,  2021）。

在本期《无创心电学年鉴》中，Younis 等人。( 2023 ) 报告称，计算遗传变异预测工具可以识别先天性长 QT 综合征 (LQTS) 1-3 的大多数致病变异。作者还发现计算分数并不能预测临床结果。

虽然令人鼓舞的是，本研究中变异预测工具与致病性相关，但同样重要的是要注意变异致病性的确定包括多种类型的证据，包括变异预测证据。重要的是，单独的计算机预测因子不应用于对变异的致病性进行分类，但可以用作遗传变异分类的证据之一。ACMG/AMP 建议指定使用计算预测因子是使用标准 PP3 和 BP4 来“支持”或反对致病性的证据水平（Care 等人，  2017）。支持级证据必须与其他更实质性的证据相结合才能对变体进行分类。此外，Pejaver 等人最近的手稿。( 2022 ) 为重新定义如何使用计算工具使用 ACMG/AMP 框架的贝叶斯适应性提供支持或反对变异致病性的证据提供了证据。这项工作表明，这些工具可以提供比支持证据更强有力的证据，并且计算工具达到这些证据水平的能力各不相同。这些作者还指出，选择单一工具用于 PP3/BP4 错义证据以避免结果选择中的偏差非常重要。

此外，尽管这些术语在文献中继续互换使用，但最近，临床遗传学和基因组学界已经开始区分术语变异分类和变异解释之间的差异。变异分类被定义为评估变异致病性的过程，而变异解释是指临床将基因检测结果与患者临床特征和家族史相结合以得出诊断的过程(Biesecker et al., 2018  )。这些细微差别在该领域具有挑战性，在未来的临床基因组工作中仔细定义和清楚地使用这些术语仍然很重要。

尤尼斯等人。( 2023 )还发现，计算机模拟工具无法预测临床结果，并得出结论，需要进行变异位置/功能分析才能更准确地解释风险。在变异分类过程之外，计算机计算工具的潜在用途（例如使用这些工具作为临床严重程度的指标）尚不清楚。对 LQTS 和其他遗传性心血管疾病的风险解释进行准确评估仍然具有挑战性，并且由于表达性可变和外显率不完全而变得复杂（Lankaputhra 和 Voskoboinik，2021  ）。除了可能有助于澄清相关基因中具有致病性和可能致病性变异的个体之间的风险分层之外，还可以使用进一步的基于人群和基因型优先的方法来进一步评估与变异解释相关的这些复杂问题。 Wilczewski 等人，  2023）。