Fig. 3

a The percentual decrease in the number of candidate links per GCF is shown for each dataset. Boxes are drawn from the first to the third quartile, separated by the median. Whiskers are extended to 1.5 times the interquartile range. b Histogram showing the number of candidate links per GCF in the Streptomyces/Salinispora dataset after co-occurrence scoring (standardised Metcalf), and after NPClassScore filtering with a cut-off of 0.25. The bin size is 25. The triangles and stars depict the number of links for the GCFs of staurosporine and rosamicin, respectively, as shown in d. c Summary for the detection of the experimentally validated links in the three datasets as listed on the PoDP. It is indicated whether links were correctly retained, incorrectly discarded due to a low standardised Metcalf score while passing the NPClassScore threshold of 0.25, or incorrectly discarded due to NPClassScore. Some validated links could not be detected as the reported spectrum on the PoDP was lacking in the dataset. Strepto/Sali is short for the Streptomyces/Salinispora dataset. d Depiction of two experimentally validated BGC-MS/MS links, for staurosporine and rosamicin, from the PoDP that are present in the dataset. The staurosporine-encoding BGC NC_009953.1.region013 from Salinispora arenicola CNS205 is shown as representative for GCF 534, linked to spectrum 89513. The rosamicin-encoding BGC NZ_AUGH01000019.region001 from Salinispora pacifica CNS237, which is fragmented due to being located on a contig edge, is shown as representative for GCF 944, linked to spectrum 130529. NPClassScore is depicted for both validated links as well as their ranks before and after filtering with NPClassScore, where OrgHetCyc is short for Organoheterocyclic compounds. Additionally, the total number of candidate MS/MS spectrum links are given for the staurosporine-and rosamicin-encoding GCF, denoted after the slash, before and after NPClassScore filtering. The number of links for the GCFs of staurosporine and rosamicin before and after NPClassScore filtering is also shown in b using the triangles and stars, respectively