Merge pull request #35 from uclahs-cds/jarbet-rerun-full-ML-pipeline

Rerun entire ML pipeline and update models

Author: jarbet

DESCRIPTION

@@ -1,11 +1,11 @@
 Package: PrCaMethy
 Title: Prostate Cancer Methylation
 Version: 1.1.0
-Date: 2025-10-07
-Maintainer: Jaron Arbet <[email protected]>
+Date: 2025-10-09
+Maintainer: Jaron Arbet <[email protected]>
 Description: Resources for predicting clinical and molecular features using prostate cancer DNA methylation data.
 Authors@R: c(
-    person("Jaron Arbet", role = c("aut", "cre"), email = "[email protected]"),
+    person("Jaron Arbet", role = c("aut", "cre"), email = "[email protected]"),
     person("Paul C. Boutros", role = "aut"))
 Depends:
     R (>= 3.5)

NEWS.md

@@ -1,8 +1,9 @@
-## PrCaMethy 1.1.0 (2025-10-07)
+## PrCaMethy 1.1.0 (2025-10-09)
 
 ### New Features
 
 - Added a new model to `all.models` for predicting `T2E.fusion`.  Thus `estimate.features()` will now predict `T2E.fusion`.
+   + For devs: previously the new T2E model and old model results were created/saved in 2 different files, making maintenance more difficult.  Thus scripts were consolidated so that all models are now created/saved in a single file.  This will improve maintenance of the package.
 
 ## PrCaMethy 1.0.0 (2025-05-22)
 

data/all.models.rda

@@ -4,14 +4,15 @@ test.mode <- FALSE;
 
 load(arg$path.ml.res);
 final.models <- readRDS(arg$path.final.models);
+res.date <- regmatches(arg$path.ml.res, regexpr('\\d{4}-\\d{2}-\\d{2}', arg$path.ml.res));
 
 outcomes <- unique(ml.res.params$outcome);
 # remove continuous psa because of cohort bias and remove categorical age since unnecessary (can just use continuous age)
 outcomes <- outcomes[!outcomes %in% c('log2.psa.continuous', 'age.categorical')];
 stopifnot(length(outcomes) == 15);
 
 if (test.mode) {
-    outcomes <- c('age.continuous', 't.stage');
+    outcomes <- c('age.continuous', 't.category');
     }
 
 ####
@@ -59,11 +60,6 @@ models <- lapply(
             ml.res.params$outcome == outcome &
             ml.res.params$top.features == final.model$top.features
             );
-        if (outcome == 'T2E.fusion') {
-            res.date <- arg$date.t2e.model;
-        } else {
-            res.date <- arg$date.old.models;
-            }
 
         file <- file.path(dirname(arg$path.ml.res), paste0(res.date, '_F72-predict-clinical-and-drivers_discrete-methyFALSE_models-', mod.id, '-', outcome, '-', final.model$top.features, '.RData'));