Table 3.

Performance metrics of the LLNA and GPMT against each another. Metrics taking into account the 10 additional records which had three conflicting outcomes are shown in parenthesis. An example of a record with three conflicting results would be a compound with sensitising and non-sensitising LNNA results and a non-sensitising GPMT result.

	Accuracy	Balanced Accuracy	Sensitivity	Specificity
LLNA is predictive of GPMT	86% (79%)	86% (79%)	93% (86%)	80% (72%)
GPMT is predictive of LLNA	86% (79%)	86% (80%)	80% (75%)	92% (84%)