2026-10-14T06:31:13Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33814278 2026-09-15T17:57:14Z category_12 category_13 category_14 category_734 category_132 category_133 portal_5 item_type_3 month_year_09_2026

<p>Subtype-classification performance (AUC) across conditions for basal-like and classical pancreatic ductal adenocarcinoma using a single global graph across patients using degree based module scores from Moffitt genes. The absolute correlation threshold was set to 0.2 to measure performance with the maximum number of valid degrees. In turn, retaining a larger number of lower-weight edges altered the degree-weighted scores among patients, whereas limiting the number of retained edges could reduce score variance across graph conditions because the graphs differed in their edge-weight distributions and edge counts. For the global-graph analysis, lowering the threshold allowed us to evaluate how heterogeneous connectivity and increased variation in degree-weighted patient scores affected model performance. All other experiments set the edge correlation threshold to 0.5 to isolate strongly correlated subsets of genes.</p> Alejandro Leyva (12268024) M. Khalid Khan Niazi (17818678) Cell Biology Genetics Molecular Biology Biological Sciences not elsewhere classified Infectious Diseases Plant Biology obtain network estimates improve subtype classification characterize biological pathways cancer genome atlas networks remains unclear multimodal lioness networks individualized lioness networks timing proxies tcga ). specific integration single samples rna modules rna expression replication timing morphological embeddings mean degree linear interpolation gene modules findings show every fold derived replication basal like <p>Subtype-classification performance (AUC) across conditions for basal-like and classical pancreatic ductal adenocarcinoma using a single global graph across patients using degree based module scores from Moffitt genes. The absolute correlation threshold was set to 0.2 to measure performance with the maximum number of valid degrees. In turn, retaining a larger number of lower-weight edges altered the degree-weighted scores among patients, whereas limiting the number of retained edges could reduce score variance across graph conditions because the graphs differed in their edge-weight distributions and edge counts. For the global-graph analysis, lowering the threshold allowed us to evaluate how heterogeneous connectivity and increased variation in degree-weighted patient scores affected model performance. All other experiments set the edge correlation threshold to 0.5 to isolate strongly correlated subsets of genes.</p> 2026-09-15T17:57:03Z Dataset Dataset 10.1371/journal.pone.0358246.t002 https://figshare.com/articles/dataset/_p_Subtype-classification_performance_AUC_across_conditions_for_basal-like_and_classical_pancreatic_ductal_adenocarcinoma_using_a_single_global_graph_across_patients_using_degree_based_module_scores_from_Moffitt_genes_The_absolute_correlati/33814278 CC BY 4.0