CompaRNA - on-line benchmarks of RNA structure prediction methods
Home

Methods
Datasets
Rankings
RNA 2D Atlas

Help
FAQ

Contact us
RSS feeds
Twitter

Table of contents:

  1. Overview

  2. Performance Plots

  3. Performance of TurboFold(20) - scored higher in this pairwise comparison

  4. Performance of NanoFolder - scored lower in this pairwise comparison

  5. Compile and download dataset for TurboFold(20) & NanoFolder [.zip] - may take several seconds...


Overview

Metric TurboFold(20) NanoFolder
MCC 0.608 > 0.271
Average MCC ± 95% Confidence Intervals 0.597 ± 0.147 > 0.294 ± 0.128
Sensitivity 0.513 > 0.286
Positive Predictive Value 0.724 > 0.266
Total TP 255 > 142
Total TN 82015 > 81834
Total FP 136 < 428
Total FP CONTRA 8 < 51
Total FP INCONS 89 < 340
Total FP COMP 39 > 37
Total FN 242 < 355
P-value 2.24270395587e-08

^top




Performance plots


  1. Comparison of performance of TurboFold(20) and NanoFolder. Positive Predictive Value (PPV) is plotted against sensitivity. Each dot represents a single test of each method. See tables below for raw data (individual counts for TurboFold(20) and NanoFolder).

  2. Average Matthews Correlation Coefficients (MCC) with 95% confidence intervals (CIs) were plotted for different RNA families, for which at least 3 members were present in the benchmarking dataset. 'n' denotes the number of MCCs used to calculate the average and CI. See tables below for raw data (individual counts for TurboFold(20) and NanoFolder).

  3. Comparison of average Matthews Correlation Coefficients (MCCs) for TurboFold(20) and NanoFolder. The whiskers correspond to 95% confidence intervals (CIs). 'n' denotes the number of MCCs used to calculate average MCCs and CIs. See tables below for raw data (individual counts for TurboFold(20) and NanoFolder).

^top





Performance of TurboFold(20) - scored higher in this pairwise comparison

1. Total counts & total scores for TurboFold(20)

Total Base Pair Counts
Total TP 255
Total TN 82015
Total FP 136
Total FP CONTRA 8
Total FP INCONS 89
Total FP COMP 39
Total FN 242
Total Scores
MCC 0.608
Average MCC ± 95% Confidence Intervals 0.597 ± 0.147
Sensitivity 0.513
Positive Predictive Value 0.724
Nr of predictions 13

^top



2. Individual counts for TurboFold(20) [ download as .csv ]

RNA Chain Rfam family MCC SENS PPV TP TN FP FP CONTRA FP INCONS FP COMP FN
3J20_1 0.96 0.91 1.00 21 2905 0 0 0 0 2
3J2L_3 0.74 0.58 0.94 31 7842 5 0 2 3 22
3J3D_C 0.82 0.71 0.95 20 2754 1 0 1 0 8
3J3E_7 0.65 0.54 0.78 29 7103 8 1 7 0 25
3J3E_8 0.17 0.15 0.21 5 7479 28 2 17 9 28
3J3F_8 0.38 0.36 0.39 13 12213 34 2 18 14 23
3J3F_7 0.79 0.68 0.92 34 7223 4 0 3 1 16
3ZEX_D 0.76 0.63 0.91 31 6987 3 0 3 0 18
3ZND_W 0.43 0.39 0.47 9 2984 12 1 9 2 14
4A1C_3 0.69 0.57 0.84 31 7103 6 0 6 0 23
4A1C_2 0.18 0.15 0.21 5 11757 28 0 19 9 28
4AOB_A 0.56 0.40 0.77 17 4349 6 1 4 1 25
4ENC_A 0.65 0.47 0.90 9 1316 1 1 0 0 10

^top



Performance of NanoFolder - scored lower in this pairwise comparison

1. Total counts & total scores for NanoFolder

Total Base Pair Counts
Total TP 142
Total TN 81834
Total FP 428
Total FP CONTRA 51
Total FP INCONS 340
Total FP COMP 37
Total FN 355
Total Scores
MCC 0.271
Average MCC ± 95% Confidence Intervals 0.294 ± 0.128
Sensitivity 0.286
Positive Predictive Value 0.266
Nr of predictions 13

^top



2. Individual counts for NanoFolder [ download as .csv ]

RNA Chain Rfam family MCC SENS PPV TP TN FP FP CONTRA FP INCONS FP COMP FN
3J20_1 0.41 0.48 0.37 11 2896 19 5 14 0 12
3J2L_3 0.11 0.11 0.12 6 7824 46 3 42 1 47
3J3D_C 0.70 0.71 0.69 20 2746 9 2 7 0 8
3J3E_7 0.35 0.33 0.37 18 7091 31 1 30 0 36
3J3E_8 0.00 0.00 0.00 0 7466 47 6 31 10 33
3J3F_8 0.27 0.33 0.22 12 12192 54 9 33 12 24
3J3F_7 0.17 0.18 0.17 9 7208 43 3 40 0 41
3ZEX_D 0.26 0.27 0.27 13 6972 36 1 35 0 36
3ZND_W 0.18 0.22 0.17 5 2973 29 6 19 4 18
4A1C_3 0.48 0.46 0.51 25 7091 24 2 22 0 29
4A1C_2 0.00 0.00 0.00 0 11728 61 12 41 8 33
4AOB_A 0.33 0.29 0.39 12 4340 20 1 18 1 30
4ENC_A 0.57 0.58 0.58 11 1307 9 0 8 1 8

^top


Matthews Correlation Coeffient, Sensitivity and Positive Predictive Value have been calculated based on the paper by Gardener & Giegerich, 2004.