Feat survival task inclusion by tonyliang19 · Pull Request #116 · CompBio-Lab/MESSI-pipeline

tonyliang19 · 2026-09-29T14:20:53Z

No description provided.

Major change to prepare data, seed reproduce

feat: add the final samplesheet used for the pipeline

…temp var option

…cli and config

…ight classifiers

…funs

…d to sklearn logistic regressions only

Copilot

Copilot review overview

🟡 Changes recommended

Survival execution currently has unresolved workflow imports, configuration gaps, incompatible result handling, and failing MOFA code paths.

Review effort: Balanced
Findings: 5 · 2

Open (7)

Route survival models to survival-specific prediction metrics · New
Include skip_sksurv in the Python-run predicate · New
Import SKSURV into the subworkflow · New
Define survival_model_names configuration · New
Add survival-aware result merging and metrics · New
Wire survival column parameters through the workflow · New
Implement single-modality Logit-only behavior · New

What changed in this PR

Adds survival-analysis support and single-modality benchmarking while reorganizing container configuration and removing deprecated simulation methods.

Changes:

Adds survival splitting, scikit-survival workflows, and preliminary MOFA survival support.
Adds unimodal dataset expansion and sklearn dimensionality reduction.
Centralizes container selection and removes deprecated simulation, GOAT, and SGMR code.

File	Description
`workflows/messi_benchmark.nf`	Routes unimodal datasets through splitting and validation.
`subworkflows/splitting/main.nf`	Passes outcome type to splitting.
`subworkflows/simulation/main.nf`	Removes the simulation workflow.
`subworkflows/prepare_data/main.nf`	Produces unimodal MuData channels.
`subworkflows/methods/sksurv/main.nf`	Adds the scikit-survival workflow.
`subworkflows/methods/sklearn/main.nf`	Adds configurable dimensionality reduction.
`subworkflows/methods/sgmr/main.nf`	Removes the SGMR placeholder.
`subworkflows/cross_validation/r/main.nf`	Removes SGMR registration.
`subworkflows/cross_validation/python/main.nf`	Routes sklearn and survival methods.
`subworkflows/cross_validation/main.nf`	Adds unimodal cross-validation inputs.
`nextflow.config`	Adds survival, sklearn, and container settings.
`modules/split_train_test/resources/usr/bin/split_tr_te.py`	Adds survival-event stratification.
`modules/split_train_test/main.nf`	Passes outcome type to the splitter.
`modules/sksurv/train/resources/usr/bin/sksurv_train.py`	Implements survival-model training.
`modules/sksurv/train/resources/usr/bin/manual_set_seed.py`	Adds reproducibility utilities.
`modules/sksurv/train/resources/usr/bin/load_tr_te.py`	Loads survival fold datasets.
`modules/sksurv/train/resources/usr/bin/load_survival_model_class.py`	Defines survival model builders.
`modules/sksurv/train/resources/usr/bin/combine_mdata2df_survival.py`	Converts MuData to survival inputs.
`modules/sksurv/train/main.nf`	Defines the survival training process.
`modules/sksurv/preprocess/resources/usr/bin/tr_te_split_mdata.py`	Partitions MuData folds.
`modules/sksurv/preprocess/resources/usr/bin/sksurv_preprocess.py`	Creates survival train/test files.
`modules/sksurv/preprocess/resources/usr/bin/load_test_splits.py`	Loads fold index files.
`modules/sksurv/preprocess/main.nf`	Defines survival preprocessing.
`modules/sksurv/predict/resources/usr/bin/sksurv_predict.py`	Generates survival predictions.
`modules/sksurv/predict/resources/usr/bin/manual_set_seed.py`	Adds prediction seed utilities.
`modules/sksurv/predict/resources/usr/bin/load_survival_model_class.py`	Supplies prediction-time model classes.
`modules/sksurv/predict/resources/usr/bin/generate_result_table_survival.py`	Formats survival results.
`modules/sksurv/predict/resources/usr/bin/combine_mdata2df_survival.py`	Converts prediction MuData inputs.
`modules/sksurv/predict/main.nf`	Defines survival prediction.
`modules/sklearn/train/resources/usr/bin/sklearn_train.py`	Adds per-modality PCA pipelines.
`modules/sklearn/train/resources/usr/bin/load_classifier_class.py`	Removes the local classifier loader.
`modules/sklearn/train/resources/usr/bin/combine_mdata2df.py`	Removes the local MuData converter.
`modules/sklearn/train/main.nf`	Adds reduction variants to training.
`modules/sklearn/select_feature/resources/usr/bin/sklearn_select_features.py`	Uses shared conversion and fitted search pipelines.
`modules/sklearn/select_feature/resources/usr/bin/run_random_search_cv.py`	Searches over scaled pipelines.
`modules/sklearn/select_feature/resources/usr/bin/combine_mdata2df.py`	Removes duplicated conversion code.
`modules/sklearn/select_feature/main.nf`	Uses centralized container labels.
`modules/sklearn/preprocess/main.nf`	Uses centralized container labels.
`modules/sklearn/predict/resources/usr/bin/sklearn_predict.py`	Adapts to the shared converter API.
`modules/sklearn/predict/resources/usr/bin/combine_mdata2df.py`	Removes duplicated conversion code.
`modules/sklearn/predict/main.nf`	Uses centralized container labels.
`modules/simulation/simulate_mvn_data/resources/usr/bin/unique_matrices.R`	Removes MVN simulation helpers.
`modules/simulation/simulate_mvn_data/resources/usr/bin/simulate_data.R`	Removes MVN simulation logic.
`modules/simulation/simulate_mvn_data/resources/usr/bin/gen_simul_metadata.R`	Removes simulation metadata generation.
`modules/simulation/simulate_mvn_data/resources/usr/bin/debug_simulate_data.R`	Removes simulation debugging code.
`modules/simulation/simulate_mvn_data/main.nf`	Removes the MVN process.
`modules/simulation/simulate_intersim/resources/usr/bin/simulate_InterSIM.R`	Removes InterSIM generation.
`modules/simulation/simulate_intersim/resources/usr/bin/save_mudata.py`	Removes InterSIM MuData serialization.
`modules/simulation/simulate_intersim/main.nf`	Removes the InterSIM process.
`modules/rgcca/train/resources/usr/bin/run_rgcca.R`	Selects the Horst scheme.
`modules/rgcca/train/main.nf`	Uses centralized RGCCA containers.
`modules/rgcca/select_feature/resources/usr/bin/rgcca_select_features.R`	Uses the Horst scheme for selection.
`modules/rgcca/select_feature/main.nf`	Uses centralized RGCCA containers.
`modules/rgcca/preprocess/main.nf`	Uses centralized RGCCA containers.
`modules/rgcca/predict/main.nf`	Uses centralized RGCCA containers.
`modules/prepare_data/uncompress_record/main.nf`	Uses the generic container label.
`modules/prepare_data/split_modality/resources/usr/bin/split_modalities.py`	Splits MuData by modality.
`modules/prepare_data/split_modality/main.nf`	Defines modality splitting.
`modules/prepare_data/prepare_mu_data/main.nf`	Uses the generic container label.
`modules/prepare_data/prepare_mae_data/main.nf`	Uses the generic container label.
`modules/prepare_data/parse_metadata/main.nf`	Uses the generic container label.
`modules/mogonet/train/main.nf`	Uses centralized MOGONET containers.
`modules/mogonet/select_feature/main.nf`	Uses centralized MOGONET containers.
`modules/mogonet/preprocess/main.nf`	Uses centralized MOGONET containers.
`modules/mogonet/predict/main.nf`	Uses centralized MOGONET containers.
`modules/mofa/train/resources/usr/bin/run_mofa_survival.R`	Adds preliminary MOFA Cox training.
`modules/mofa/train/main.nf`	Selects training scripts by outcome.
`modules/mofa/select_feature/main.nf`	Uses centralized MOFA containers.
`modules/mofa/preprocess/resources/usr/bin/preprocess_mofa.R`	Documents survival embedding use.
`modules/mofa/preprocess/main.nf`	Uses centralized MOFA containers.
`modules/mofa/predict/main.nf`	Uses centralized MOFA containers.
`modules/merge_selected_features/main.nf`	Uses the CODIA container label.
`modules/merge_result_table/main.nf`	Uses the CODIA container label.
`modules/local/samplesheet_check/main.nf`	Uses the MOGONET container label.
`modules/integrao/train/main.nf`	Uses centralized Integrao containers.
`modules/integrao/select_feature/main.nf`	Uses centralized Integrao containers.
`modules/integrao/preprocess/main.nf`	Uses centralized Integrao containers.
`modules/integrao/predict/main.nf`	Uses centralized Integrao containers.
`modules/goat/main.nf`	Removes the GOAT placeholder.
`modules/diablo/train/resources/usr/bin/tune_diablo.R`	Uses the Horst tuning scheme.
`modules/diablo/train/resources/usr/bin/run_diablo.R`	Uses the Horst training scheme.
`modules/diablo/train/main.nf`	Uses centralized CODIA containers.
`modules/diablo/select_feature/resources/usr/bin/diablo_select_features.R`	Uses Horst for feature selection.
`modules/diablo/select_feature/main.nf`	Uses centralized CODIA containers.
`modules/diablo/preprocess/main.nf`	Uses centralized CODIA containers.
`modules/diablo/predict/main.nf`	Uses centralized CODIA containers.
`modules/diablo/downstream/main.nf`	Uses centralized CODIA containers.
`modules/cooperative_learning/train/main.nf`	Uses centralized CODIA containers.
`modules/cooperative_learning/select_feature/main.nf`	Uses centralized CODIA containers.
`modules/cooperative_learning/preprocess/main.nf`	Uses centralized CODIA containers.
`modules/cooperative_learning/predict/main.nf`	Uses centralized CODIA containers.
`modules/caret_multimodal/train/main.nf`	Uses centralized caret containers.
`modules/caret_multimodal/select_feature/main.nf`	Uses centralized caret containers.
`modules/caret_multimodal/preprocess/main.nf`	Uses centralized caret containers.
`modules/caret_multimodal/predict/main.nf`	Uses centralized caret containers.
`modules/calculate_metrics/resources/usr/bin/calculate_metrics.py`	Expands classification metrics.
`modules/calculate_metrics/main.nf`	Uses label-based container selection.
`launcher_local.sh`	Adds a local pipeline launcher.
`conf/real_data.config`	Corrects the feature-selection parameter name.
`bin/python_utils/load_classifier_class.py`	Centralizes and expands classifier definitions.
`bin/python_utils/combine_mdata2df.py`	Centralizes MuData conversion.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

+		// Dynamically determine script based on classification or survival task
+		def script_names = params.outcome_type == "classification" ? "run_mofa.R" : "run_mofa_survival.R"


 			params.skip_mogonet && 
-			params.skip_integrao
+			params.skip_integrao &&
+			params.skip_sklearn


-include { GOAT } 								from "${subworkflowDir}/methods/goat"
+include { INTEGRAO  }             from "${subworkflowDir}/methods/integrao"
+include { SKLEARN   }             from "${subworkflowDir}/methods/sklearn"
+include { MOGONET   } 						from "${subworkflowDir}/methods/mogonet"


+
+workflow SKSURV {
+  // Survival models to train: coxnet | rsf | gbm
+  model_name  = Channel.fromList(params.survival_model_names)


+                      [ it[2], it[3] ]
+                    }
+                    .set { result_table }
+      MERGE_RESULT_TABLE ( result_table, saveMode )


+    time_col            = "time"    // colData column with follow-up time in days
+    event_col           = "status"  // colData column with event indicator (0/1)


+  // Set classifier to Logit only if single_mode is true
  model_name = Channel.fromList(params.sklearn_classifier_names)


tonyliang19 added 30 commits February 27, 2026 19:11

Merge pull request #110 from CompBio-Lab/dev

138cd5a

Major change to prepare data, seed reproduce

Merge pull request #111 from CompBio-Lab/dev

85b2b54

feat: add the final samplesheet used for the pipeline

fix: move container option to process level config, and remove stale …

e308292

…temp var option

fix: moved all container definition into config

b0c6ba5

fix: changed wrong syntax, adding args to pass in splitting modules, …

09c4905

…cli and config

fix: lint conf syntax, included back sklearn

0819287

rm: remove old files

5f7cd2e

rm: remove redundant staled configs

0ac61ac

fix: rename metric labels

2da5e23

rm: remove deprecated workflows

939ac4c

feat: sklearn to use pca50 or no reduction for more models

7944abd

fix: update method options

ec7e648

fix: include select feature for sklearn models

e5ef195

fix: correct sklearn reduction to list param rather than singleton

7c99113

fix: correct naming of parameter opt

6f0ee70

fix: make sure sklearn cv search works even with make_pipeline

0ce67f6

fix: use eipy container for sklearn as it contains xgboost, and use r…

b66e821

…ight classifiers

fix: moved common sklearn utils to bin/python_utils, and standardize …

e190ae8

…funs

fix: refactor code for sklearn select feature

e0956f7

feat: update sklearn random search cv params as roc

44547e9

fix: finalize models in sklearn

36a5963

feat: add new option of sklearn single modality

09fbe25

feat: include functionality to handle single modality running, limite…

eaefa01

…d to sklearn logistic regressions only

feat: include missing scripts of splitting modalities

43753b9

fix: coerce scheme of rgcca and diablo to 'horst'

9245afc

fix: make sure modality names are valid when exporting to filenames

1fbaaeb

feat: add new boolean to run classification or survival

06ede3a

fix: remove old skipper of methods

8f4a27e

feat: add sksurv modules (not tested, need to add container def too

b598a38

feat: mofa try survival, need to run later

6fbc86f

feat: this THU specific launcher local script wrapper, del later

bb719ea

Copilot AI balanced review requested due to automatic review settings September 29, 2026 14:20

Copilot started reviewing on behalf of tonyliang19 September 29, 2026 14:21 View session

Copilot AI reviewed Sep 29, 2026

View reviewed changes

tonyliang19 merged commit 0423fcc into dev Sep 29, 2026
1 check passed

Navigation Menu

Repository navigation

Feat survival task inclusion - #116

Feat survival task inclusion#116
tonyliang19 merged 31 commits into
devfrom
feat-survival-task-inclusion

tonyliang19 commented Sep 29, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		// Dynamically determine script based on classification or survival task
		def script_names = params.outcome_type == "classification" ? "run_mofa.R" : "run_mofa_survival.R"

		time_col = "time" // colData column with follow-up time in days
		event_col = "status" // colData column with event indicator (0/1)

		// Set classifier to Logit only if single_mode is true
		model_name = Channel.fromList(params.sklearn_classifier_names)

Conversation

tonyliang19 commented Sep 29, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants