Run
All the tables and the summary figure in one zip file.
Download all (zip)
The probability is how sure the classifier is of the subtype it assigns. Below 50% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 70% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 33% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 30% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 30% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 70% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 70% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 70% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
The probability is how sure the classifier is of the subtype it assigns. Below 70% the sample is flagged as low confidence: read it as undetermined rather than as belonging to that subtype. Subtype colours are the ones used in the paper.
Tumor and stroma proportions
Estimated fraction of epithelium (E), stroma (S) and other cells (O) of each sample, from the virtual microdissection. Only filled in when stroma classification is included.
Tumor classification
Collisson
Moffitt
Bailey
Puleo
Chan-Seng-Yue
PDAConsensus
Stroma classification
Moffitt
Maurer
PDAConsensus
Number of samples assigned to each subtype, in the colours of the paper. Filled in when a classification finishes.
Mean balanced accuracy of each classifier, as published (Genome Medicine 2025, Tables 4 and 6), estimated by repeated 10-times 10-fold cross-validation on the training cohort of 514 samples. The dashed shape is what a classifier that always predicts the most frequent class would reach.
Tumor classifiers
Stroma classifiers
Input file
Format
Raw counts in a tab-separated (.tsv) or comma-separated (.csv) file: gene IDs in the first column, one column per sample, up to 1 GB.
Gene IDs
Ensembl IDs may include the version, which is removed. Rows that end up sharing a gene ID are averaged.
Batch correction
Your samples are corrected together with the 514 training samples, treating yours as one batch. The result depends on which samples are uploaded together, so use the same setting for the whole study.
How long it takes
About 5 minutes for 5 samples and 20 minutes for 100, with batch correction. Stroma classification roughly doubles the time.
Reading the results
Probability
How sure the classifier is of the subtype it assigns. Below the threshold of its classifier the sample is flagged as low confidence and should be read as undetermined.
- 70% for Moffitt, the stroma classifiers and PDAConsensus
- 50% for Collisson
- 33% for Bailey
- 30% for Puleo and Chan-Seng-Yue
Scores
NonClassicalScore and ActivatedECMScore are the probability of belonging to the non-classical and activated-ECM classes, as continuous values between 0 and 1.
About
Version: v2.6.4
Cite the method: Villoslada-Blanco P, Alonso L, Sabroso-Lasa S, Maquedano M, Estudillo L, Real FX, López de Maturana E, Malats N. Development of a consensus molecular classifier for pancreatic ductal adenocarcinoma. Genome Medicine 2025;17:142. 10.1186/s13073-025-01568-9
Cite the software: 10.5281/zenodo.17019896
Code: github.com/pavillos/PDACMOC
Contact: pvilloslada@cnio.es
Use and citations
Citations
Downloads
This app
License
