Integrating Computational Lead Optimization Diagnostics With Analog Design and Candidate Selection

Abstract

Aim: Combining computational lead optimization diagnostics with analog design and computational approaches for assessing optimization efforts are discussed and the compound optimization monitor is introduced. Methods: Approaches for compound potency prediction are described and a new analog design algorithm is introduced. Calculation protocols are detailed. Results & discussion: The study rationale is explained. Compound optimization monitor diagnostics are combined with a thoroughly evaluated approach for compound design and candidate prioritization. The diagnostic scoring scheme is further extended. Future perspective: Opportunities for practical applications of the integrated computational methodology are described and further development perspectives are discussed.

DeepCOMO: from structure-activity relationship diagnostics to generative molecular design using the compound optimization monitor methodology

“…Furthermore, the choice of chemical reference spaces for compound distance calculations is variable and can be modified according to the characteristic features and requirements of specific optimization efforts. Herein, a seven-dimensional chemical reference space composed of seven LO-relevant physicochemical descriptors (calculated with RDKit [ 21 ]) was used, which was shown to provide sufficient chemical resolution for the characterization of ASs in our previous studies [ 3 – 5 ]. Since the NBH concept plays a central role in the COMO approach, score calculations depend on the definition of a suitable NBH radius that adequately mirrors distances between EAs and VAs.…”

“…In addition to AS scaffold-based enumeration, the Free-Wilson (FW) additivity principle [ 22 , 23 ] has been adapted and converted into a design strategy for generating VA candidates for late LO stages [ 5 ]. Therefore, matched molecular pairs (MMPs) [ 24 ] are calculated for EAs of an AS and systematically organized in MMP networks (generated with the NetworkX Python library [ 25 ]).…”

“…Ensuing compound quartets meeting FW requirements consist of three EAs and an FW EA whose putative potency is predicted based upon FW principles. Such quartets represent local mini-QSAR models that have been shown to be surprisingly accurate in many cases and capable of complementing global QSAR strategies for VA prediction and prioritization [ 5 ]. Accuracy of FW predictions intrinsically depends on the presence of SAR continuity.…”

“…One of the roots of COMO was the development of a scoring scheme to evaluate chemical saturation of compound series on the basis of biological screening data [ 1 , 2 ]. Modifying and extending this scoring scheme and combining it with the assessment of SAR progression then gave rise to the introduction of the COMO approach [ 3 – 5 ], which was originally designed as a diagnostic. On the basis of COMO scoring, ASs can be assigned to different development stages.…”

“…Thus, although COMO was originally devised as a diagnostic/descriptive tool it also had the intrinsic potential to bridge between LO analysis and compound design. Accordingly, different analog design strategies and activity prediction approaches have been implemented in COMO to design and prioritize VAs [ 5 , 6 ].…”