Table 2. Benchmarking performance improvement for netDx.

Computation times are averaged over four runs of the same ten queries for feature-scoring one patient label, while limiting the executable to a single core. All tests were performed on an Intel Xeon @ 2.6GHz machine with 126GB of available RAM and 12 cores.

JavaMemory setting	Previous runtime (s) v0.99	Current runtime (s) with percent improvement v1.4
Breast cancer (Luminal A): 111 patients, 1706 pathway-based networks from gene expression data
4GB	273.95 +/- 13.97	167.73 +/- 5.21 (38%)
6GB	275.11 +/- 12.35	166.89 +/- 4.81 (39%)
8GB	273.74 +/- 13.44	167.24 +/- 4.44 (39%)
Schizophrenia (case): 185 patients, 1735 pathway-based networks from gene expression data
4GB	552.41 +/- 26.06	389.37 +/- 11.26
6GB	549.31 +/- 23.83	388.22 +/- 9.14
8GB	547.47 +/- 21.93	391.85 +/- 10.93