Derivatization Design of Synthetically Accessible Space for Optimization: In Silico Synthesis vs Deep Generative Design

Abstract

Molecular design is of utmost importance in lead optimization programs ultimately determining the fate of the project and the speed to reach preclinical stage. Newly designed lead analogues or new chemotypes must successfully address the challenges in the multidimensional optimization process throughout several optimization cycles. The speed, quality, and creativity of the designs can have a major impact on the cycle time, the number of required cycles, and the number of compounds needed to be synthesized and evaluated that in combination affect the overall timeline and cost of the lead optimization phase. Recently, a new concept, generative design with deep learning, has become popular for de novo design of project relevant analogue sets. We have developed a de novo design technology called "derivatization design" that applies artificial-intelligence-assisted forward in silico synthesis for the generation of near neighbor lead analogues as well as scaffold variations. The several attractive features of the methodology include synthetic feasibility, reagent availability and cost data associated with each new molecule; thus, detailed synthetic assessment is automatically generated during the design. As a result, these practically important data types can become an early part of the ranking and selection process for cycle time reduction. The power of derivatization design is demonstrated in a simple design study of DDR1 inhibitors and comparison of the produced molecules to a recently published data set obtained with deep generative design.

UnCorrupt SMILES: a novel approach to de novo design

“…SMILES exploration resulted in compounds with similar bioactivities to already known compounds, but with a higher density for bioactivities around 6.0 log units, whilst the target-directed RNN was able to produce more compounds with higher predicted activities (Figure 5D). However, previous studies on derivatization design have suggested that there are also merits to staying closer to known actives, as in the case of Makara et al this does lead to a higher hit rate compared to a target-directed VAE [61].…”

CoPriNet: graph neural networks provide accurate and rapid compound price prediction for molecule prioritisation

“…However, as described above, SAscore and SCscore are simple approximations for SA and as such, present several limitations. For instance, it is well known that these scores tend to underestimate the SA of difficult compounds that can be synthesized from complex commercially available building blocks 35,36 . It has also been shown that structurally similar compounds, which tend to have similar complexity-based scores, can require synthetic strategies of different difficulty levels 35 , leading to incorrect SA estimations.…”

CoPriNet: Deep learning compound price prediction for use in de novo molecule generation and prioritization.

“…However, as described above, SAscore and SCscore are simple approximations for SA and as such, present several limitations. For instance, it is well known that these scores tend to underestimate the SA of difficult compounds that can be synthesized from complex commercially available building blocks (Gao and Coley, 2020;Makara et al, 2021). It has also been shown that structurally similar compounds, which also tend to have similar complexitybased scores, can require synthetic strategies of different difficulty levels (Gao and Coley, 2020), leading to incorrect SA estimations.…”