Computational Drug Discovery and Design

group matching information for an assayed molecule with the

reference molecule DKPES (seeNote 2).

Please note that this protocol assumes that a tabular dataset

containing information on the molecules as well as the assay

response has already been collected. However, the analysis

approach outlined in this chapter is a general one, and it is not

restricted to the specific functional group matching patterns shown

in Fig.6. For more information on how this functional group

matching data can be generated from a ligand-based screening,see

[10](https://github.com/psa-lab/screenlamp).

The first column of the DKPES data table (Fig.6), “index,”

numbers each molecule. The “Signal Inhibition” column contains

the response variable measured by the biological assay, in this case

ranging from 0 (non-active) to 1 (highly active, with 100% DKPES

signal inhibition). For instance, we can see from the table (Fig.6)

that ENE4 and ZINC72400307 (petromyzonol sulfate) were the

most promising candidate inhibitors, as they reduced the olfactory

response to DKPES by 90.5% and 90.4%, respectively, when each

inhibitor candidate was used at the same equimolar concentration

(10
^6 M) as DKPES. The consequent columns, labeled as 3-Keto,

3-Hydroxy, and so forth, contain information about whether an

atom or functional group in the candidate molecule overlayed

Fig. 6Code for reading the DKPES dataset into a data frame. The characters>>>denote a Python interpreter
prompting for a command to enter and execute. The table resulting from the execution of this code example
(df.head(10)) shows an excerpt from the DKPES data table sorted by signal inhibition: the ten most
active molecules from the EOG experiments and their functional group matching patterns

316 Sebastian Raschka et al.