Sachem: a chemical cartridge for high-performance substructure search

Table 1 Overview of indexing and searching errors

Cartridge	Indexing failures			Rejected queries
Cartridge	1M	10M	94M	n	reason
Bingo	105	1024	9754	0
OrChem	2	4	–	12	Unsupported aromatic bond in SMILES
pgchem	30	255	2527	146	Fragmented SMILES, queries with [*]
RDKit	72	707	6911	4	Chemical structure considered invalid
Sachem	0	0	0	0

Measurements are slightly influenced by errors that some cartridges exhibited during benchmarking, due to both indexing and searching errors. Indexing errors are primarily reported as unacceptable data in the SDF files from PubChem, most frequently as invalid atom valences or stereochemistry. Note that Bingo beta version can lower the number of indexing errors by using algorithms that work with ‘incorrect’ structures (this feature is disabled by default)

ISSN: 1758-2946