A molecular classification of diseases that accurately reflects clinical behaviour lays the foundation of precision medicine. The development of in silico classifiers coupled with molecular implementation based on DNA reactions marks a key advance in more powerful molecular classification, but it nevertheless remains a challenge to process multiple molecular datatypes. Here we introduce a DNA-encoded molecular classifier that can physically implement the computational classification of multidimensional molecular clinical data. To produce unified electrochemical sensing signals across heterogeneous molecular binding events, we exploit DNA-framework-based programmable atom-like nanoparticles with n valence to develop valence-encoded signal reporters that enable linearity in translating virtually any biomolecular binding events to signal gains. Multidimensional molecular information in computational classification is thus precisely assigned weights for bioanalysis. We demonstrate the implementation of a molecular classifier based on programmable atom-like nanoparticles to perform biomarker panel screening and analyse a panel of six biomarkers across three-dimensional datatypes for a near-deterministic molecular taxonomy of prostate cancer patients.

准确反映临床行为的疾病分子分类奠定了精准医疗的基础。计算机分类器的发展与基于 DNA 反应的分子实现相结合，标志着更强大的分子分类的关键进步，但处理多种分子数据类型仍然是一个挑战。这里我们介绍一种DNA编码的分子分类器，它可以物理地实现多维分子临床数据的计算分类。为了在异质分子结合事件中产生统一的电化学传感信号，我们利用基于 DNA 框架的n价可编程原子状纳米粒子来开发价编码信号报告器，使几乎任何生物分子结合事件都能线性地转换为信号增益。因此，计算分类中的多维分子信息被精确地分配权重以用于生物分析。我们演示了基于可编程类原子纳米粒子的分子分类器的实现，以执行生物标志物面板筛选并分析跨三维数据类型的六个生物标志物面板，以对前列腺癌患者进行近乎确定性的分子分类。