XXIX National Meeting on Medicinal Chemistry

Schrödinger is excited to be participating in the XXIX National Meeting on Medicinal Chemistry conference taking place on September 14th – 17th in Parma, Italy. Join us for a presentation by Giulia D’Arrigo, Senior Scientis I at Schrödinger, titled “Accelerating Drug Discovery through Integrated Physics-Based and Machine Learning Approaches using Schrödinger’s Computational Platform.”

Speaker:

Giulia D’Arrigo, Senior Scientis I, Schrödinger

Abstract:

Schrödinger’s computational platform accelerates molecular design and drug discovery by integrating physics-based methods and machine learning to address the complexities of multi-parameter optimization using a comprehensive suite of in silico tools and workflows spamming the entire research and development pipeline. Here, real-world applications of these technologies will be illustrated across different case studies from the Schrödinger Therapeutics Group (STG).

Ultra-large scale chemical space exploration using AutoDesigner [1,2] allows the generation and efficient prioritization of potentially billions of novel structures to systematically explore novel chemical scaffolds or R-groups that meet project requirements. Combining machine learning with rigorous free energy calculations using FEP+ [3,4] in an iterative fashion (Active Learning FEP+), further allows to rapidly score the ultra-large libraries generated by AutoDesigner to identify designs with optimal potency and selectivity profile. Examples of this automated large scale approach are the identification, after 7 months of project initiation, of novel WEE1 inhibitors with >10,000X selectivity over PLK [1] as well as the discovery of four novel scaffolds of EFGR inhibitors in only six days [5].

A new, in-silico approach for addressing ADME/T liabilities and enabling precise control over key endpoints is presented. This approach leverages state of the art computational tools such as induced-fit docking (IFD-MD) [6,7], FEP+, and solvation energy calculations with E-sol [8], in order to enable a rational approach to ADME. Applications of these workflows to hERG inhibition, CYP rate of metabolism, and brain exposure and efflux are presented. . A prime example is the in silico-enabled discovery of KAI-11101 [9], a DLK inhibitor for neurodegenerative disease treatment. Here, accurate ADMET profiling and large scale on-target and off-target FEP+ were fundamental to elect KAI-11101 as a brain penetrant, potent and highly selective kinase inhibitor preclinical candidate.

Altogether, these examples demonstrate the impact of Schrödinger’s computational platform in facilitating and accelerating drug discovery programs.

References

[1] Bos et al. J Chem Inf Model. 2024 Oct 14;64(19):7513-7524.
[2] Bos et al. J. Chem. Inf. Model. 2022, 62, 8, 1905–1915.
[3] Wang et al.  J. Am. Chem. Soc., 2015, 137(7), 2695–2703.
[4] Ross et al. Commun. Chem., 2023, 6(222).
[5] Igawa et al. J Med Chem. 2024 Dec 26;67(24):21811-21840.
[6] Miller EB, et al. Cell, 2024, 187, 3, 521-525.
[7] Miller et al. J. Chem. Theory Comput. 2021, 17, 4, 2630–2639.
[8] Lawrenz et al. J Chem Inf Model. 2023 Jun 26;63(12):3786-3798.
[9]Lagiakos et al. J Med Chem. 2025 Feb 13;68(3):2720-2741.