Skip to main content

Table 1 Results from simulation studies using data generated from the proposed methodology. The value of F for HBFM represents the number of factors in the “best” model choice, as determined by DIC

From: A sparse Bayesian factor model for the construction of gene co-expression networks from single-cell RNA sequencing count data

  Sim 1: N=125,Fsim=10      Sim 2: N=125,Fsim=15
  TPR FDR AUC Edges   TPR FDR AUC Edges
HBFM, F = 15 0.760 0.153 0.927 314 HBFM, F = 15 0.640 0.111 0.820 306
LEAP 0.386 0.378 0.705 217 LEAP 0.341 0.275 0.665 200
PIDC 0.634 0.293 0.821 314* PIDC 0.506 0.297 0.742 306*
SCODE 0.229 0.745 0.550 314* SCODE 0.249 0.654 0.504 306*
BN 0.206 0.077 0.682 78 BN 0.186 0.037 0.672 82
GENIE3 0.540 0.398 0.746 314* GENIE3 0.468 0.350 0.711 306*
PCORR 0.123 0.566 0.599 99 PCORR 0.148 0.442 0.602 113
  Sim 3: N=500,Fsim=10      Sim 4: N=500,Fsim=15
  TPR FDR AUC Edges   TPR FDR AUC Edges
HBFM, F = 25 0.889 0.034 0.984 322 HBFM, F = 25 0.704 0.029 0.929 308
LEAP 0.743 0.608 0.741 664 LEAP 0.402 0.305 0.696 246
PIDC 0.794 0.137 0.915 322* PIDC 0.621 0.143 0.866 308*
SCODE 0.249 0.730 0.501 322* SCODE 0.216 0.701 0.578 308*
BN 0.277 0.040 0.751 101 BN 0.212 0.032 0.716 93
GENIE3 0.554 0.398 0.754 322* GENIE3 0.466 0.357 0.729 308*
PCORR 0.300 0.266 0.683 143 PCORR 0.261 0.327 0.624 165
  Sim 5: N=1000,Fsim=10      Sim 6: N=1000,Fsim=15
  TPR FDR AUC Edges   TPR FDR AUC Edges
HBFM, F = 20 0.909 0.076 0.973 344 HBFM, F = 25 0.624 0.070 0.904 285
LEAP 0.780 0.550 0.804 606 LEAP 0.591 0.541 0.680 547
PIDC 0.857 0.128 0.954 344* PIDC 0.633 0.056 0.889 285*
SCODE 0.269 0.727 0.496 344* SCODE 0.221 0.670 0.510 285*
BN 0.323 0.050 0.793 119 BN 0.247 0.037 0.710 109
GENIE3 0.603 0.387 0.764 344* GENIE3 0.440 0.344 0.700 285*
PCORR 0.403 0.291 0.720 199 PCORR 0.294 0.251 0.669 167
  1. *Number of edges fixed to match HBFM