函数内limma::removeBatchEffect报错，外部运行正常问题排查

我编写了一个函数，通过布尔型参数use_batch控制是否执行批次效应校正，函数内的校正代码块如下：

if (use_batch){
  scores = as.data.frame(scores)
  cancer.type = metadata[['Cancer_Type']]
  scores = limma::removeBatchEffect(t(scores) ,batch = cancer.type)
  scores = t(scores)
}

运行该函数时出现报错：

Error in rowMeans(y$exprs, na.rm = TRUE) : 'x' must be numeric

但直接用真实数据my_scores和my_meta在函数外执行完全相同的逻辑却正常运行：

> if (TRUE){
     scoresx = as.data.frame(my_scores)
     cancer.type = my_meta[['Cancer_Type']]
     scoresx = limma::removeBatchEffect(t(scoresx) ,batch = cancer.type)
     scoresx = t(scoresx)
 }

注意：当use_batch=FALSE时函数运行完全正常，回溯显示错误出在scores = limma::removeBatchEffect(t(scores) ,batch = cancer.type)这一行。

以下是数据子集：

> dput(my_scores[1:6,1:6])
structure(c(-0.0741098696855045, -0.094401270881699, 0.0410284948786532, 
-0.163302950330185, -0.0942478217207681, -0.167314411991775, 
-0.178835447722468, -0.253897294559596, -0.0372301980787381, 
-0.230579110769457, -0.224125346052727, -0.196933050675633, -0.0384421660291032, 
-0.0275306107582565, 0.186447606591857, -0.124972070102036, -0.15348122673842, 
-0.106812144494277, -0.0924083910597563, -0.172356328661097, 
-0.0172673823614314, 0.0280649471541352, -0.128925304635747, 
-0.0875076743713435, -0.0680848706469295, -0.173427291586957, 
-0.0106773958944477, -0.0015805672257001, -0.0751114943036091, 
-0.0737177243152751, -0.0391833488213571, -0.0275279418713283, 
0.0156454755097513, 0.0285160860867748, -0.0633367938488132, 
0.0252778805872529), dim = c(6L, 6L), dimnames = list(c("Pt1", 
"Pt10", "Pt101", "Pt103", "Pt106", "Pt11"), c("CD4-T-cells", 
"CD8-T-cells", "T-helpers", "NK-cells", "Monocytes", "Neutrophils"
)))

> dput(my_meta[1:6,2:7])
structure(list(samples = c("Pt1", "Pt10", "Pt101", "Pt103", "Pt106", 
"Pt11"), Response = c("NoResponse", "NoResponse", "Response", 
"NoResponse", "NoResponse", "NoResponse"), RECIST = c("PD", "SD", 
"PR", "PD", "PD", "PD"), Gender = c("male", "female", "female", 
"female", "male", "female"), treatment = c("anti-PD1", "anti-PD1", 
"anti-PD1", "anti-PD1", "anti-PD1", "anti-PD1"), Cancer_Type = c("Melanoma", 
"Melanoma", "Melanoma", "Melanoma", "Melanoma", "Melanoma")), row.names = c("Pt1", 
"Pt10", "Pt101", "Pt103", "Pt106", "Pt11"), class = "data.frame")

• 变量作用域不匹配：函数内部的metadata可能未正确绑定到输入的元数据参数，而是引用了全局环境中其他非预期对象，导致cancer.type的长度/结构与scores转置后的样本数不匹配；或者函数内的scores被其他代码修改，转置后不再是纯数值结构。而全局环境的my_scores和my_meta是明确匹配的，因此运行正常。
• 数据类型识别偏差：limma::removeBatchEffect会优先将data.frame识别为ExpressionSet对象（这也是报错中出现y$exprs的原因），而非普通数值矩阵。函数内转置后的scores可能因结构问题被误判为非数值矩阵，导致函数尝试访问ExpressionSet专属的exprs属性，但实际对象无此属性，进而引发rowMeans处理非数值对象的错误。而全局环境中t(scoresx)是纯数值矩阵，函数会按普通矩阵处理，因此无问题。
• 隐式类型转换问题：函数内的scores转换为data.frame后，可能混入了非数值元素（比如字符型列），转置后整个对象变成字符型，不满足removeBatchEffect对数值输入的要求。而全局环境的my_scores是纯数值矩阵，转置后仍保持数值型，因此能正常运行。

• 在函数内明确将scores转换为数值矩阵：替换scores = as.data.frame(scores)为scores = as.matrix(scores)，确保转置后是纯数值矩阵再传入removeBatchEffect。
• 检查函数定义，确保metadata作为参数传入，避免依赖全局变量，保证批次变量与样本一一对应。
• 验证cancer.type的长度与t(scores)的行数（即原scores的列数）完全一致，避免维度不匹配问题。

内容的提问来源于stack exchange，提问作者Kev