Bigras Gilbert, Dong Wei-Feng, Canil Sarah, Lai Raymond, Morel Didier, Swanson Paul E, Izevbaye Iyare
Departments of Laboratory Medicine and Pathology, Cross Cancer Institute.
Becton Dickinson, Office of Science, Medicine and Technology, Corporate Clinical Development, Pont-de-Claix, France.
Appl Immunohistochem Mol Morphol. 2018 Jan;26(1):54-63. doi: 10.1097/PAI.0000000000000367.
A new automated MYC IHC classifier based on bivariate logistic regression is presented. The predictor relies on image analysis developed with the open-source ImageJ platform. From a histologic section immunostained for MYC protein, 2 dimensionless quantitative variables are extracted: (a) relative distance between nuclei positive for MYC IHC based on euclidean minimum spanning tree graph and (b) coefficient of variation of the MYC IHC stain intensity among MYC IHC-positive nuclei. Distance between positive nuclei is suggested to inversely correlate MYC gene rearrangement status, whereas coefficient of variation is suggested to inversely correlate physiological regulation of MYC protein expression. The bivariate classifier was compared with 2 other MYC IHC classifiers (based on percentage of MYC IHC positive nuclei), all tested on 113 lymphomas including mostly diffuse large B-cell lymphomas with known MYC fluorescent in situ hybridization (FISH) status. The bivariate classifier strongly outperformed the "percentage of MYC IHC-positive nuclei" methods to predict MYC+ FISH status with 100% sensitivity (95% confidence interval, 94-100) associated with 80% specificity. The test is rapidly performed and might at a minimum provide primary IHC screening for MYC gene rearrangement status in diffuse large B-cell lymphomas. Furthermore, as this bivariate classifier actually predicts "permanent overexpressed MYC protein status," it might identify nontranslocation-related chromosomal anomalies missed by FISH.
本文介绍了一种基于二元逻辑回归的新型自动化MYC免疫组化分类器。该预测器依赖于使用开源ImageJ平台开发的图像分析。从免疫染色检测MYC蛋白的组织切片中，提取两个无量纲定量变量：(a)基于欧几里得最小生成树图的MYC免疫组化阳性细胞核之间的相对距离，以及(b)MYC免疫组化阳性细胞核中MYC免疫组化染色强度的变异系数。阳性细胞核之间的距离被认为与MYC基因重排状态呈负相关，而变异系数被认为与MYC蛋白表达的生理调节呈负相关。将该二元分类器与其他2种MYC免疫组化分类器（基于MYC免疫组化阳性细胞核百分比）进行比较，所有分类器均在113例淋巴瘤上进行测试，其中大多为已知MYC荧光原位杂交（FISH）状态的弥漫性大B细胞淋巴瘤。该二元分类器在预测MYC+FISH状态方面明显优于“MYC免疫组化阳性细胞核百分比”方法，灵敏度为100%（95%置信区间，94-100），特异性为80%。该检测执行迅速，至少可为弥漫性大B细胞淋巴瘤的MYC基因重排状态提供初步免疫组化筛查。此外，由于该二元分类器实际上预测的是“MYC蛋白持续过表达状态”，它可能识别出FISH遗漏的与非易位相关的染色体异常。
