Table 1.

Effect of missing spectra in the model input.

metric		full dataset	4 spectra	3 spectra	2 spectra	1 spectrum
# test cases		1000	413	65	483	39
Avg. MW		275.3	287.5	242.6	267.4	300.3
Avg. SMILES length		34.5	37.0	28.5	32.5	43.6
correct molecules (↑)	(%)	7.0	9.2	15.2	4.1	5.1
correct formulas (↑)	(%)	39.3	45.1	46.9	34.8	20.5
DMW% (↓)	Min	2.3	1.6	0.5	2.4	9.5
	Avg	6.3	5.5	3.9	6.6	14.6
DMF% (↓)	Min	9.2	6.5	8.1	10.8	21.1
	Avg	21.7	17.8	24.5	24.0	32.9
Fngpcosine (↑)	Max	0.53	0.56	0.57	0.50	0.45
	Avg	0.36	0.39	0.38	0.34	0.31
MCSratio (↑)	Max	0.68	0.70	0.72	0.66	0.57
	Avg	0.51	0.53	0.55	0.50	0.43
${\mathrm{MCS}}_{\tan }$ (↑)	Max	0.55	0.58	0.60	0.53	0.44
	Avg	0.38	0.39	0.41	0.36	0.30
MCScoef (↑)	Max	0.71	0.73	0.74	0.69	0.63
	Avg	0.54	0.55	0.58	0.53	0.48

Evaluation metrics when considering the entire test set and the test-data partitions that have available all 4, only 3, only 2, and only 1 spectrum. The arrows show the desired trend for each metric.