# Improving the SSIR Method: Implementation of an Exhaustive Multilevel Scan for Categorical Variables

## TL;DR

This paper improves a chemical analysis method to better identify molecular patterns linked to drug activity.

## Contribution

The paper introduces a multilevel scan extension to the SSIR method for analyzing molecular substituents.

## Key findings

- The extended SSIR method can handle groups of residues in addition to single residues.
- The new framework generalizes the original SSIR method as a special case.
- The approach is demonstrated with anti-HIV compounds and virtual molecular predictions.

## Abstract

The Superposing Significant Interaction Rules (SSIR) method is briefly revisited. The original version of the algorithm consists of dealing with rules that evaluate the individual presence or absence of substituents in the molecular structures of a combinatorial chemical family. Here, an extension is developed, allowing for the systematic consideration of multilevel conditions, i.e., the presence or absence of not only single residues but also groups of residues attached to a particular substitution site. This possibility expands the universe of possible definable rules. Then, the original SSIR version becomes a particular case of the framework presented in this manuscript. A numerical example involving a family of anti-HIV compounds is provided to illustrate the application and potential of the extended method. Procedures for generating predictions for new in silico virtual molecular structures are also shown.