Skip to main content
Fig. 3 | BMC Neuroscience

Fig. 3

From: NCX-DB: a unified resource for integrative analysis of the sodium calcium exchanger super-family

Fig. 3

Comparison of NCX-DB exchanger classification to simulated random data. The classification feature of NCX-DB was benchmarked against a simulated dataset. In the NCX-DB filtering algorithm, NCX-DB searches for alpha repeat structures and filters candidates based upon log odds scores obtained for each motif. To benchmark this filtering algorithm, we compared the prediction accuracy of NCX-DB for a training dataset of known sodium calcium exchangers to a randomized data set. 100, 000 randomized alpha repeats were generated using a custom Perl script (Additional file 1). The density of log odds scores for each randomized alpha repeat is shown as histograms. a NCX alpha 1 repeat randomized data. 95th percentile = 4.3, 99th percentile = 8.6 coefficient of variation (CV) = 9.84; c NCX alpha 2 repeat randomized data. 95th percentile = 6.4, 99th percentile = 8.6, CV = 1.5; e NCKX alpha 1 repeat randomized data. 95th percentile = 5.44, 99th percentile = 8.6, CV = 2.13; g NCKX alpha repeat 2 randomized data. 95th percentile = 5.1, 99th percentile = 8.6, CV = 2.45; i NCLX alpha repeat 1 randomized data. 95th percentile = 5.89, 99th percentile = 8.6, CV = 1.67; k NCLX alpha repeat 2 randomized data. 95th percentile = 5.99, 99th percentile = 8.4, CV = 1.43. In all cases we observed the highest density of log odds scores close to zero. The 95th percentile log odds ranged from 4.3 to 6.4, and 99th percentile log odds ranged from 8.4 to 8.6. We then used this data to examine how well NCX-DB predicts the correct exchanger in a background of randomized data. The ROC curves in each case revealed AUC values close or at 1 for each motif: b NCX alpha 1 repeat AUC = 0.979 and random NCX alpha 1 repeat AUC = 0.4744; d the NCX alpha 2 repeat AUC = 1 and random NCX alpha 2 repeat AUC = 0.4491; f NCKX alpha 1 repeat AUC = 1 and random NCKX alpha 1 repeat AUC = 0.4622; h NCKX alpha 2 repeat AUC = 1 and random NCKX alpha 2 repeat AUC = 0.4896; j NCLX alpha 1 repeat AUC = 1 and random NCLX alpha 1 repeat AUC = 0.509; l NCLX alpha 2 repeat AUC = 0.9985 and random NCLX alpha 2 repeat AUC = 0.5458. Histograms were generated using the hist(x,…) function in R. ROC curves were made in R using the ROCR package (https://rocr.bioinf.mpi-sb.mpg.de/) using the perf ← performance(pred,”tpr”,”fpr”) to plot true positive rate against false positive rate. Control randomized data was added (add = TRUE) to the experimental colorized data

Back to article page