Traditional scientific discovery has been flipped on its head, in large part because of enormous amounts of ‘omics data generated from human samples. Scientific discovery previously initiated in model systems has given way to a human-forward approach where genomic, proteomic, and metabolomic profiling (‘omics) is performed in thousands of biospecimens in just months. Such studies can facilitate mechanistic insights, identify those at greatest risk of disease, and provide new targets for intervention, but what has been the return on investment of these costly experiments in translation to biomarkers or therapeutics used for patient care?

Since the human genome was published 2 decades ago, technical advances and shrinking costs have fueled explosive growth of large-scale genetic sequencing in human cohorts. However, whereas the study of rare protein-coding genetic variants has revolutionized diagnostics for rare monogenic diseases, genetic investigation of common complex cardiovascular disease (CVD) has had a longer runway to success. The biologic elusiveness of the noncoding genome, conflicting clinical use of polygenic risk scores, and a historical focus on cohorts of European ancestry present opportunities for ongoing research, but also highlight the natural (and slow) arc of investigation. This long CVD genetics runway has resulted in breakthroughs with effects on population health. Studies of humans harboring loss-of-function sequence variations mimicking genetic knockouts resulted in development of PCSK9 (proprotein convertase subtilisin/kexin type 9) inhibitors and new therapeutics in the pipeline, including ANGPTL3 (angiopoietin-like 3) and ApoC-III (apolipoprotein C-III) inhibitors. Discovery fueled by human genetics research has also led to therapeutics targeting kidney disease caused by APOL1 sequence variations.

Whereas genetics is foundational, cardiometabolic diseases arise from a complex interplay of genetics, diet, lifestyle, and interorgan communication present for years before clinical presentation. This challenging landscape provides fertile ground for ’omics approaches, where molecular markers are measured in biospecimens from human cohorts, often with an unbiased lens. These markers are more proximal and dynamic reporters of disease processes and integrate environmental influences. ‘Omics studies leverage sequencing, mass spectrometry, and affinity-based technologies, enabling monitoring of thousands of RNA transcripts, metabolites, or proteins (transcriptomics, metabolomics, and proteomics, respectively) from small amounts of samples with rapid throughput. In the case of metabolomics (the systematic study of chemical byproducts of biologic pathways), blood profiling has identified unanticipated disease predictors including amino acids,¹ acylcarnitines,² and lipid species. Likewise, proteomic profiling has identified circulating proteins whose levels portend disease¹ orthogonal to clinical risk factors.³

These readily accessible platforms and the growing tranche of publicly available data have led to a literature rife with thousands of ‘omics disease associations and threads of mechanistic insights, but have these findings bridged the proverbial “valley of death” for translating from bench to bedside? There are few clear winners in this divide in CVD, save for AlloMap (which is diagnostic for cardiac transplant rejection). Whereas several new biomarkers are being brought to market, including new CVD protein biomarkers, their success in clinical uptake and improving cardiovascular health remains unknown. The reasons for the lack of translation of ‘omics discovery into patient care are manifold, and include difficulties in modeling complex biologic processes and a need for better bioinformatics, more implementation science funding, and cultures and infrastructures that support team science. In addition, perhaps most importantly, it takes time. The overhyped promise of personalized medicine has led to unrealistic expectations about how quickly translation occurs, particularly for CVDs, which are complex, manifesting clinically over decades.

How can ‘omics science sift through these biochemical associations to better translate from bench to bedside? For therapeutic target prioritization, human genetics remains a critical anchor. Levels of circulating ‘omics markers are heritable and genetics can explain a proportion of interindividual variability. This information can be used to examine genetic determinants of flux in pathways and test whether markers are causally related to disease, taking advantage of the natural experiment provided by the random assortment of genetic alleles (Mendelian randomization). This paradigm has been demonstrated for known genes associated with lipoprotein levels (eg, PCSK9, LDLR), but is being increasingly applied to novel biomarkers. A recent publication identified dozens of proteins associated with incident heart failure, several showing evidence of a causal relationship and druggability, including adrenomedullin.³

‘Omics cannot meet its full potential devoid of information about an individual’s clinical status. Emerging phenotyping approaches have enabled more granular dissection of disease: machine learning detects patterns in images, digital health devices provide second-to-second snapshots of physiology, and longitudinal electronic health data create roadmaps across an individual lifespan. Pairing ‘omics with these fonts of clinical information has expanded the data space for sifting while facilitating important discoveries. In concordance, cardiometabolic diseases lumped together in a reductionist approach are now recognized as more specific subtypes⁴ upon which ‘omics data are now being mapped. Such holistic approaches will likely continue to advance more personalized approaches in rare (ie, cardiomyopathy therapies targeted to specific genetic sequence variations) and common CVDs.

Retrotranslation back to the bench is mandatory. Follow-up investigations in model organisms or in stem cell models are critical to parsing the nuggets of candidates, dissecting molecular pathways, and understanding target tissues, interorgan relationships, and potential off-target effects. Advances in single-cell ‘omics also offer mechanistic interrogation of signals identified through bulk profiling of biospecimens with heterogeneous cellular composition.⁵

The ‘omics revolution has created terabytes of data to sift through, much of which is available through a few keystrokes on a computer. Our collective ability to find the biologic nuggets of gold in these data and bridge the as-of-yet mostly unbridged cardiovascular ‘omics translational device requires interdisciplinary team science leveraging clinical expertise; computational prowess to build, share, and analyze multidimensional data; careful assessment and application of emerging ‘omics technologies; mechanistic assessments in preclinical models for signals identified in human cohorts (Figure); and purposeful efforts to train, support, and inspire the next generation of scientists in this space. Whereas our successes in bridging bench to bedside (or the reverse) in CVD ‘omics can be measured on a single hand, the future holds enormous potential to prevent and treat disease and improve human health through harnessing the power of ‘omics with a careful filter.

Figure. Framework for contemporary ‘omics studies. Depicted is a framework for how ‘omics can yield mechanistic and biomarker insights: (1) ‘omics platforms are applied to large human cohorts, (2) including integrative ‘omics (ie, for identifying quantitative trait loci useful in Mendelian randomization analyses to assess whether ‘omics analytes are in the casual pathway for cardiometabolic disease); paired with (3) careful deep phenotyping including image-based deep learning; and (4) mechanistically validated in single-cell ‘omics and preclinical models, which are (5) supported by foundational computational analyses, team science, and a need to support training of the next generation of scientists.

This work was supported by the following grants: HHSN268201600034I, U24 DK112340, R01 NR019628, R01 DK081572, and R01 HL133870 (to Dr Gerszten) and R01HL146145 and R21AI158756 (to Dr Shah).

Disclosures Dr Shah receives research funding through sponsored research agreements to Duke University from AstraZeneca, Lilly Inc, Verily Inc, and nference, and is a coinventor on 2 unlicensed patents held by Duke University.

The American Heart Association celebrates its 100th anniversary in 2024. This article is part of a series across the entire AHA Journal portfolio written by international thought leaders on the past, present, and future of cardiovascular and cerebrovascular research and care. To explore the full Centennial Collection, visit https://www.ahajournals.org/centennial

The opinions expressed in this article are not necessarily those of the editors or of the American Heart Association.

For Sources of Funding and Disclosures, see page 1325.

Circulation is available at www.ahajournals.org/journal/circ

传统的科学发现已经发生了翻天覆地的变化，很大程度上是因为人类样本产生了大量的“组学数据”。以前在模型系统中启动的科学发现已经让位于人类前瞻性的方法，在短短几个月内就可以在数千个生物样本中进行基因组、蛋白质组和代谢组学分析（“组学”）。此类研究可以促进机制洞察，识别疾病风险最大的人群，并提供新的干预目标，但这些昂贵的实验转化为用于患者护理的生物标志物或疗法的投资回报是多少？

自二十年前公布人类基因组以来，技术进步和成本下降推动了人类群体大规模基因测序的爆炸性增长。然而，尽管对罕见蛋白质编码遗传变异的研究彻底改变了罕见单基因疾病的诊断，但常见复杂心血管疾病 (CVD) 的基因研究却有更长的成功之路。非编码基因组的生物学难以捉摸、多基因风险评分的临床使用相互矛盾，以及历史上对欧洲血统群体的关注，为正在进行的研究提供了机会，但也强调了自然（且缓慢）的研究弧线。这条漫长的心血管疾病遗传学跑道已经取得了突破，对人口健康产生了影响。对人类携带功能丧失序列变异的模仿基因敲除的研究导致了 PCSK9（前蛋白转化酶枯草杆菌蛋白酶/kexin 9 型）抑制剂和正在研发中的新疗法的开发，包括 ANGPTL3（血管生成素样 3）和 ApoC-III（载脂蛋白） C-III) 抑制剂。人类遗传学研究推动的发现也催生了针对APOL1序列变异引起的肾脏疾病的治疗方法。

虽然遗传学是基础，但心脏代谢疾病是由遗传学、饮食、生活方式和器官间通讯的复杂相互作用引起的，这些相互作用在临床表现之前就已经存在多年。这种具有挑战性的环境为“组学方法”提供了肥沃的土壤，在“组学方法”中，通常使用公正的镜头来测量人类群体生物样本中的分子标记。这些标记是疾病过程的更近端和动态的报告者，并整合环境影响。组学研究利用测序、质谱和基于亲和力的技术，能够快速监测少量样品中的数千种 RNA 转录本、代谢物或蛋白质（分别为转录组学、代谢组学和蛋白质组学）。就代谢组学（生物途径化学副产物的系统研究）而言，血液分析已识别出意料之外的疾病预测因子，包括氨基酸、¹酰基肉碱、²和脂质种类。同样，蛋白质组学分析已鉴定出循环蛋白质，其水平预示着与临床危险因素正交的疾病^{1 。 3}

这些易于访问的平台和不断增长的公开数据导致了充斥着数千个“组学疾病关联”和机械见解线索的文献，但这些发现是否跨越了众所周知的“死亡之谷”，从实验室转化为临床？除了 AlloMap（用于诊断心脏移植排斥反应）之外，在 CVD 领域中几乎没有明显的赢家。尽管一些新的生物标志物正在被推向市场，包括新的 CVD 蛋白质生物标志物，但它们在临床应用和改善心血管健康方面的成功仍然未知。组学发现未能转化为患者护理的原因是多方面的，包括复杂生物过程建模的困难以及对更好的生物信息学、更多实施科学资金以及支持团队科学的文化和基础设施的需求。此外，也许最重要的是，这需要时间。对个性化医疗的过度夸大的承诺导致人们对转化发生的速度产生了不切实际的期望，特别是对于心血管疾病来说，这种疾病很复杂，需要数十年的时间才能在临床上显现出来。

组学科学如何筛选这些生化关联，以更好地从实验室转化为临床？对于治疗目标的优先顺序，人类遗传学仍然是一个关键的锚点。循环组学标记的水平是可遗传的，遗传学可以解释一定比例的个体间变异。该信息可用于检查通路中流动的遗传决定因素，并利用遗传等位基因随机分类（孟德尔随机化）提供的自然实验来测试标记是否与疾病因果相关。这种范例已在与脂蛋白水平相关的已知基因（例如PCSK9、LDLR）中得到证实，但正越来越多地应用于新型生物标志物。最近的一篇出版物鉴定了数十种与心力衰竭相关的蛋白质，其中一些蛋白质显示出因果关系和可药性的证据，其中包括肾上腺髓质素。³

如果缺乏有关个人临床状态的信息，组学就无法充分发挥其潜力。新兴的表型分析方法使得对疾病进行更精细的剖析成为可能：机器学习检测图像中的模式，数字健康设备提供秒到秒的生理快照，纵向电子健康数据创建跨个体生命周期的路线图。将组学与这些临床信息字体配对，扩大了筛选的数据空间，同时促进了重要发现。一致地，以还原论方法集中在一起的心脏代谢疾病现在被认为是更具体的亚型⁴，现在正在绘制“组学数据”。这种整体方法可能会继续在罕见（即针对特定基因序列变异的心肌病治疗）和常见心血管疾病中推进更加个性化的方法。

必须将回译回替补席。模型生物体或干细胞模型的后续研究对于解析候选者的核心、剖析分子途径以及了解靶组织、器官间关系和潜在的脱靶效应至关重要。单细胞组学的进步还提供了对通过对具有异质细胞组成的生物样本进行批量分析而识别的信号进行机械询问。⁵

组学革命创造了数万亿字节的数据可供筛选，其中大部分数据只需在计算机上敲击几下键盘即可获得。我们在这些数据中找到生物金块并桥接迄今为止尚未桥接的心血管组学转化设备的集体能力需要利用临床专业知识的跨学科团队科学；构建、共享和分析多维数据的计算能力；仔细评估和应用新兴的组学技术；对人类队列中识别的信号进行临床前模型的机械评估（图）；并有目的地努力培训、支持和激励该领域的下一代科学家。尽管我们在 CVD 组学中将实验台与床边（或相反）桥接起来的成功可以用单手来衡量，但未来通过仔细过滤组学的力量，在预防和治疗疾病以及改善人类健康方面拥有巨大的潜力。

数字。 当代“组学”研究框架。描述的是“组学如何产生机制和生物标志物见解”的框架：（1）“组学平台应用于大型人类群体，（2）包括综合组学（即，用于识别孟德尔随机化分析中有用的数量性状基因座，以评估“组学分析物是否存在于心脏代谢疾病的偶然途径中”；与（3）仔细的深度表型分析相结合，包括基于图像的深度学习； (4) 在单细胞组学和临床前模型中进行机械验证，(5) 得到基础计算分析、团队科学以及支持下一代科学家培训的需求的支持。

这项工作得到了以下赠款的支持：HHSN268201600034I、U24 DK112340、R01 NR019628、R01 DK081572 和 R01 HL133870（给 Gerszten 博士）以及 R01HL146145 和 R21AI158756（给 Shah 博士）。

Shah博士通过阿斯利康 (AstraZeneca)、礼来公司 (Lilly Inc)、Verily Inc. 和 nference 与杜克大学签订的赞助研究协议获得研究经费，并且是杜克大学持有的 2 项未经许可专利的共同发明人。

美国心脏协会将于 2024 年庆祝成立 100 周年。本文是国际思想领袖撰写的整个 AHA 期刊系列文章的一部分，内容涉及心脑血管研究和护理的过去、现在和未来。要探索完整的百年纪念收藏，请访问 https://www.ahajournals.org/centennial

本文表达的观点不一定代表编辑或美国心脏协会的观点。

有关资金来源和披露信息，请参阅第 1325 页。

流通量可在 www.ahajournals.org/journal/circ 上获取