Table 2. Summary of performance measures of STEED compared with manual human ascertainment.

	Specificity	Sensitivity	Precision	Accuracy	F1-score
Training corpus (motor neuron diseases, n = 45)
Species	NA	96	100	96	0.98
Sex	67	85	94	82	0.89
Disease model	NA	96	100	96	0.98
Outcome histology	89	92	97	91	0.94
Outcome behaviour	50	97	84	84	0.90
Outcome imaging	96	NA	NA	96	NA
Randomization	84	96	89	91	0.93
Blinding	95	92	96	93	0.94
Animal welfare	NA	86	97	84	0.92
Conflict of interest	100	98	100	97	0.99
Sample size calculation	78	92	63	82	0.75
ARRIVE guidelines	100	100	100	100	1.00
Data availability	85	94	94	91	0.94
Validation corpus 1 (motor neuron diseases, n = 31)
Species	NA	100	100	100	1.00
Sex	100	74	100	84	0.85
Disease model	NA	90	100	90	0.95
Outcome histology	100	96	100	97	0.98
Outcome behaviour	78	85	76	81	0.79
Outcome imaging	NA	100	100	100	1.00
Randomization	100	86	100	97	0.92
Blinding	100	89	100	97	0.94
Animal welfare	100	89	100	90	0.94
Conflict of interest	92	94	94	94	0.94
Sample size calculation	81	80	44	81	0.57
ARRIVE guidelines	100	NA	NA	100	NA
Data availability	96	83	83	94	0.83
Validation corpus 2 (multiple sclerosis, n = 244)
Species	NA	75	100	75	0.86
Sex	76	83	93	82	0.88
Disease model	NA	87	100	88	0.93
Outcome histology	64	96	93	91	0.95
Outcome behaviour	66	91	81	82	0.86
Outcome imaging	NA	94	100	94	0.97
Randomization	93	81	75	90	0.78
Blinding	98	85	96	93	0.90
Animal welfare	86	80	95	82	0.87
Conflict of interest	96	97	90	97	0.93
Sample size calculation	94	100	27	97	0.43
ARRIVE guidelines	100	100	100	100	1.00
Data availability	100	80	80	100	0.80

Specificity, sensitivity, precision, and accuracy are denoted in percentage. For details regarding measures, please see the materials and methods section. Items reaching or exceeding our pre-defined thresholds (sensitivity of 85% and a specificity of 80%) are printed in bold font.