摘要：已有研究称，病理学基础模型能够在组织形态中编码分子程序，但相关证据通常是队列范围内的基因排序列表，而非对留出患者的预测。我们以患者为证据单位重建了这类分析，并考察究竟是流水线中的哪个组件承载了信号。
在 11 个冻结骨干模型、4 个预先指定的基因程序，以及 285 名具有配对切片和 RNA-seq 数据的 TCGA-BRCA 患者中（44 个单元；按患者进行 GroupKFold；所有预处理均在折内拟合），对均值池化嵌入使用岭回归，可预测留出患者的程序得分，Spearman rho 为 0.25–0.56；UNI2 在全部四个程序上最强（免疫程序为 0.556）。匹配的置换零假设检验显示，在 10,000 次置换下，每个单元的原始 p 值约为 1e-4；Holm 校正后的 p 值为 0.0044。
该信号真实存在，但并非一致地来自形态学信息。与在相同患者和相同交叉验证折上运行的竞争模型相比，嵌入特征在 ER/腔面型、增殖和免疫程序上优于组织组成特征（分别为 +0.280、+0.284、+0.479；p <= 0.003），但在 basal 程序上并非如此：仅使用区室比例即可达到 0.469，而嵌入特征为 0.493（p = 0.77）。54 个可解释的细胞计数特征在每个程序上的差距仅为 0.043–0.085。
几何方法没有带来任何可测量的贡献，我们也解释了原因：测地线图通过欧氏最近邻搜索选择邻居，只对已选边进行重新加权，因此其拓扑在构建时就是欧氏的（黎曼几何减欧氏几何 = +0.0010，95% CI [-0.0007, +0.0029]）。在一致应用时，几何方法表现更差（-0.0117）。岭回归优于图与度量解码器 +0.097（CI [+0.069, +0.127]）。该领域常见的驱动基因计数指标在这里几乎没有信息量：91.8% 的随机六基因面板能够恢复 >=5/6 个驱动基因。

Abstract:Pathology foundation models are reported to encode molecular programmes in tissue morphology, but the evidence is usually a cohort-wide ranked gene list rather than a prediction for a held-out patient. We rebuild such an analysis with the patient as the unit of evidence and ask which pipeline component carries signal.
Across 11 frozen backbones, four pre-specified gene programmes and 285 TCGA-BRCA patients with paired slides and RNA-seq (44 cells; GroupKFold by patient, all preprocessing fitted inside the fold), ridge regression on mean-pooled embeddings predicts held-out programme scores at Spearman rho = 0.25-0.56, UNI2 strongest on all four (immune 0.556). A matched permutation null gives raw p ~ 1e-4 at 10,000 permutations for every cell; Holm-adjusted p = 0.0044.
The signal is real but not uniformly morphological. Against competing models on the same patients and folds, embeddings beat tissue composition for ER/luminal, proliferation and immune (+0.280, +0.284, +0.479; p <= 0.003) but not basal, where compartment fractions alone reach 0.469 against the embedding's 0.493 (p = 0.77). Fifty-four interpretable cell-count features come within 0.043-0.085 on every programme.
The geometric machinery contributes nothing measurable, and we identify why: the geodesic graph selects neighbours by Euclidean nearest-neighbour search and only reweights edges already chosen, so the topology is Euclidean by construction (Riemannian minus Euclidean = +0.0010, 95% CI [-0.0007, +0.0029]). Applied consistently the geometry is worse (-0.0117). Ridge regression beats the graph-and-metric decoder by +0.097 (CI [+0.069, +0.127]). The driver-count metric common in this literature is near-uninformative here: 91.8% of random six-gene panels recover >=5/6 drivers.