Selected article for: "test set and training set"

Author: Delli Ponti, Riccardo; Marti, Stefanie; Armaos, Alexandros; Tartaglia, Gian Gaetano
Title: A high-throughput approach to profile RNA structure
  • Document date: 2017_3_17
  • ID: k23xlzj0_12
    Snippet: Since one technique might not be sufficient to capture structural properties of long transcripts (14) , we evaluated different approaches to combine the five CROSS models (PARS-Human, PARS-Yeast, SHAPE-HIV, icSHAPE-Mouse and NMR/X-ray) into a consensus prediction. To this aim, we measured the performances of the models on an independent test set of 67 NMR/X-ray structures (15) for which SHAPE data are available (17 145 fragments in total), evalua.....
    Document: Since one technique might not be sufficient to capture structural properties of long transcripts (14) , we evaluated different approaches to combine the five CROSS models (PARS-Human, PARS-Yeast, SHAPE-HIV, icSHAPE-Mouse and NMR/X-ray) into a consensus prediction. To this aim, we measured the performances of the models on an independent test set of 67 NMR/X-ray structures (15) for which SHAPE data are available (17 145 fragments in total), evaluating precision (PPV) and Area Under the ROC Curve (AUC; Supplementary Figure S3 ). Consistently with the type of information contained in the training set, we observed the best performances for the NMR/X-ray model (PPV: 0.69; AUC: 0.64) followed by HIV-SHAPE (PPV: 0.64; AUC: 0.63). Comparing the scores of the five models, we did not find strong correlations (Supplementary Table S2), except for PARS-Human and PARS-Yeast (Pearson's correlation = 0.50) that were trained on data obtained with the same experimental techniques (Figure 1 ; Supplementary Figures S1 and S2 ).

    Search related documents:
    Co phrase search for related documents
    • consensus prediction and experimental technique: 1
    • consensus prediction and good performance: 1
    • consensus prediction and model score: 1
    • consensus prediction and Pearson correlation: 1
    • CROSS model and different approach: 1, 2
    • CROSS model and good performance: 1, 2, 3
    • CROSS model and model performance: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25
    • CROSS model and model performance measure: 1, 2
    • CROSS model and model score: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11
    • different approach and experimental technique: 1, 2
    • different approach and information type: 1
    • different approach and model performance: 1, 2, 3, 4, 5, 6, 7, 8
    • different approach and model score: 1, 2, 3
    • experimental technique and model performance: 1
    • good performance and information type: 1
    • good performance and model performance: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25
    • good performance and model score: 1, 2, 3, 4, 5, 6, 7, 8, 9
    • good performance and Pearson correlation: 1, 2, 3, 4, 5, 6
    • information type and model performance: 1