Skip to main content

Table 2 Results from simulation studies using data generated from the NORTA algorithm. 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 7: N=125, EdgesSim=350      Sim 8: N=125, EdgesSim=425
  TPR FDR AUC Edges   TPR FDR AUC Edges
HBFM, F = 18 0.817 0.043 0.927 299 HBFM, F = 18 0.638 0.000 0.942 271
LEAP 0.700 0.377 0.819 393 LEAP 0.466 0.423 0.630 343
PIDC 0.714 0.164 0.875 299* PIDC 0.560 0.122 0.817 271*
SCODE 0.243 0.716 0.491 299* SCODE 0.169 0.734 0.540 271*
BN 0.211 0.026 0.718 76 BN 0.148 0.000 0.689 63
GENIE3 0.349 0.592 0.576 299* GENIE3 0.306 0.520 0.558 271*
PCORR 0.157 0.396 0.559 91 PCORR 0.132 0.434 0.564 99
  Sim 9: N=500, EdgesSim=350      Sim 10: N=500, EdgesSim=425
  TPR FDR AUC Edges   TPR FDR AUC Edges
HBFM, F = 22 0.871 0.041 0.993 318 HBFM, F = 20 0.727 0.019 0.947 315
LEAP 0.880 0.548 0.909 681 LEAP 0.744 0.557 0.741 713
PIDC 0.857 0.057 0.941 318* PIDC 0.722 0.025 0.933 315*
SCODE 0.249 0.726 0.571 318* SCODE 0.240 0.676 0.520 315*
BN 0.251 0.064 0.764 94 BN 0.195 0.057 0.669 88
GENIE3 0.366 0.597 0.601 318* GENIE3 0.355 0.521 0.569 315*
PCORR 0.220 0.280 0.609 107 PCORR 0.198 0.408 0.571 142
  Sim 11: N=1000, EdgesSim=350      Sim 12: N=1000, EdgesSim=425
  TPR FDR AUC Edges   TPR FDR AUC Edges
HBFM, F = 25 0.966 0.048 0.989 355 HBFM, F = 25 0.831 0.033 0.968 365
LEAP 0.926 0.623 0.870 859 LEAP 0.849 0.529 0.809 767
PIDC 0.957 0.056 0.977 355* PIDC 0.807 0.060 0.940 365*
SCODE 0.303 0.701 0.506 355* SCODE 0.191 0.778 0.625 365*
BN 0.343 0.084 0.756 131 BN 0.233 0.075 0.734 107
GENIE3 0.414 0.592 0.596 355* GENIE3 0.374 0.564 0.565 365*
PCORR 0.297 0.373 0.654 166 PCORR 0.228 0.276 0.612 134
  1. *Number of edges fixed to match HBFM