1. scone
  2. Commit 633b426fe80b901a525bc92aa81b117b9de93067
Browse code

EXP_.._COR methods now measure R^2

mbcole authored on 13/09/2017 23:48:47
Showing 2 changed files
R/scone_eval.R
History View file @ 633b426
... ... 
@@ -63,11 +63,11 @@
63 63 
 #'   silhouette width by biological condition.} \item{BATCH_SIL}{ Average
64 64 
 #'   silhouette width by batch condition.} \item{PAM_SIL}{ Maximum average
65 65 
 #'   silhouette width from PAM clustering (see stratified_pam argument).}
66 
-#'   \item{EXP_QC_COR}{ Maximum squared spearman correlation between expression
66 
+#'   \item{EXP_QC_COR}{ Coefficient of determination between expression
67 67 
 #'   pcs and quality factors (see stratified_cor argument).} \item{EXP_UV_COR}{
68 
-#'   Maximum squared spearman correlation between expression pcs and negative
68 
+#'   Coefficient of determination between expression pcs and negative
69 69 
 #'   control gene factors (see stratified_cor argument).} \item{EXP_WV_COR}{
70 
-#'   Maximum squared spearman correlation between expression pcs and positive
70 
+#'   Coefficient of determination between expression pcs and positive
71 71 
 #'   control gene factors (see stratified_cor argument).} \item{RLE_MED}{ The
72 72 
 #'   mean squared median Relative Log Expression (RLE) (see stratified_rle
73 73 
 #'   argument).} \item{RLE_IQR}{ The variance of the inter-quartile range (IQR)
... ... 
@@ -268,23 +268,26 @@ score_matrix <- function(expr,
268 268
269 269 
       # Max cor with quality factors.
270 270 
       if (!is.null(qc_factors)) {
271 
-        EXP_QC_COR <- EXP_QC_COR + cond_w * max(cor(proj[is_cond, ],
272 
-                                                    qc_factors[is_cond, ],
273 
-                                                    method = "spearman") ^ 2)
271 
+        EXP_QC_COR <-
272 
+          EXP_QC_COR + cond_w * (1 - sum(unlist(apply(proj, 2, function(y) {
273 
+            lm(y ~ qc_factors)$residual
274 
+          })) ^ 2) / sum(scale(proj, scale = FALSE) ^ 2))
274 275 
       }
275 276
276 277 
       # Max cor with UV factors.
277 278 
       if (!is.null(uv_factors)) {
278 
-        EXP_UV_COR  <- EXP_UV_COR + cond_w * max(cor(proj[is_cond, ],
279 
-                                                     uv_factors[is_cond, ],
280 
-                                                     method = "spearman") ^ 2)
279 
+        EXP_UV_COR  <-
280 
+          EXP_UV_COR + cond_w * (1 - sum(unlist(apply(proj, 2, function(y) {
281 
+            lm(y ~ uv_factors)$residual
282 
+          })) ^ 2) / sum(scale(proj, scale = FALSE) ^ 2))
281 283 
       }
282 284
283 285 
       # Max cor with WV factors.
284 286 
       if (!is.null(wv_factors)) {
285 
-        EXP_WV_COR <-  EXP_WV_COR + cond_w * max(cor(proj[is_cond, ],
286 
-                                                     wv_factors[is_cond, ],
287 
-                                                     method = "spearman") ^ 2)
287 
+        EXP_WV_COR <-
288 
+          EXP_WV_COR + cond_w * (1 - sum(unlist(apply(proj, 2, function(y) {
289 
+            lm(y ~ wv_factors)$residual
290 
+          })) ^ 2) / sum(scale(proj, scale = FALSE) ^ 2))
288 291 
       }
289 292
290 293 
     }
... ... 
@@ -308,21 +311,27 @@ score_matrix <- function(expr,
308 311 
   } else{
309 312 
     # Max cor with quality factors.
310 313 
     if (!is.null(qc_factors)) {
311 
-      EXP_QC_COR = max(cor(proj, qc_factors, method = "spearman") ^ 2)
314 
+      EXP_QC_COR <- 1 - sum(unlist(apply(proj, 2, function(y) {
315 
+        lm(y ~ qc_factors)$residual
316 
+        })) ^ 2) / sum(scale(proj, scale = FALSE) ^ 2)
312 317 
     } else{
313 318 
       EXP_QC_COR = NA
314 319 
     }
315 320
316 321 
     # Max cor with UV factors.
317 322 
     if (!is.null(uv_factors)) {
318 
-      EXP_UV_COR = max(cor(proj, uv_factors, method = "spearman") ^ 2)
323 
+      EXP_UV_COR <- 1 - sum(unlist(apply(proj, 2, function(y) {
324 
+        lm(y ~ uv_factors)$residual
325 
+      })) ^ 2) / sum(scale(proj, scale = FALSE) ^ 2)
319 326 
     } else{
320 327 
       EXP_UV_COR = NA
321 328 
     }
322 329
323 330 
     # Max cor with WV factors.
324 331 
     if (!is.null(wv_factors)) {
325 
-      EXP_WV_COR = max(cor(proj, wv_factors, method = "spearman") ^ 2)
332 
+      EXP_WV_COR <- 1 - sum(unlist(apply(proj, 2, function(y) {
333 
+        lm(y ~ wv_factors)$residual
334 
+      })) ^ 2) / sum(scale(proj, scale = FALSE) ^ 2)
326 335 
     } else{
327 336 
       EXP_WV_COR = NA
328 337 
     }
man/score_matrix.Rd
History View file @ 633b426
... ... 
@@ -65,11 +65,11 @@ A list with the following metrics: \itemize{ \item{BIO_SIL}{ Average
65 65 
   silhouette width by biological condition.} \item{BATCH_SIL}{ Average
66 66 
   silhouette width by batch condition.} \item{PAM_SIL}{ Maximum average
67 67 
   silhouette width from PAM clustering (see stratified_pam argument).}
68 
-  \item{EXP_QC_COR}{ Maximum squared spearman correlation between expression
68 
+  \item{EXP_QC_COR}{ Coefficient of determination between expression
69 69 
   pcs and quality factors (see stratified_cor argument).} \item{EXP_UV_COR}{
70 
-  Maximum squared spearman correlation between expression pcs and negative
70 
+  Coefficient of determination between expression pcs and negative
71 71 
   control gene factors (see stratified_cor argument).} \item{EXP_WV_COR}{
72 
-  Maximum squared spearman correlation between expression pcs and positive
72 
+  Coefficient of determination between expression pcs and positive
73 73 
   control gene factors (see stratified_cor argument).} \item{RLE_MED}{ The
74 74 
   mean squared median Relative Log Expression (RLE) (see stratified_rle
75 75 
   argument).} \item{RLE_IQR}{ The variance of the inter-quartile range (IQR)