The identification of 158 compounds missed by a naïve similarity search demonstrates the complementary chemical space sampled in a hit explosion setup.
An example of a compound missing an assignment is the experimental drug cytidine-5′-diphospho-beta-delta-xylose (CID46936568), which was only classified as a pyrimidine ribonucleoside diphosphate but not classified as a purine nucleotide sugar (Fig. 9b).
After removing compounds with missing structure information, the dataset contains 42 689 compounds, 1504 of which are classified as active.
In those reactions where one compound was missing, the known compounds were compared to each other.
We delete 327 compounds with missing or inconsistent annotations of the ion mode, as well as 254 compounds for which we do not find a PubChem identifier, and 201 compounds that are not fully connected by covalent bonds.
A k-step reaction sequence is defined as a series of reactions in which a chemical compound is known to be converted to another compound by multiple enzymatic reactions, and the corresponding (k– 1) intermediate compounds are missing.
This is critical, as several biochemical pathways for key compounds are missing enzymatic activities, and in cases where entire pathways are missing, a known transporter for a particular compound is also absent.
