23rd Schrödinger European User Group Meeting

Additional talk information and the detailed agenda will be added as the event approaches.

Dominik Schwarz, Bayer 

In silico variant screening facilitates resistance mutation identification 

Drug resistance is a major issue in cancer therapy. Knowledge about resistance mutations could prevent unnecessary patient treatment with ineffective drugs, in clinical trials as well as clinical practice, or potentially speed-up the development of follow-up compounds. Therefore, we evaluated on-target resistance mutation identification with in silico methods, namely free energy perturbation (FEP) affinity estimates in combination with protein fitness estimates. We compared our predictions with deep mutational scanning (DMS) data that tested for resistance caused by single amino acid mutations. Our results show that protein fitness estimates are useful for increasing the precision in resistance mutation identification.

Dylan Serillon, WhiteLab Genomics

Revolutionizing peptide discovery: AI-driven drug design paving the way for the future of cell & gene therapy 

Over the past decade, rational peptide design has gained significant momentum as a promising approach in drug discovery. Peptides are emerging as highly versatile therapeutic agents, offering unique advantages in targeting complex biological pathways. However, their design and development still represent a significant challenge due to their intrinsic structural flexibility and potential for unintended biological interactions, which can complicate therapeutic applications. Beyond their therapeutic potential, peptides may enhance gene delivery through various vector types and modifications. For Adeno-Associated Virus (AAVs) vectors, peptides can be inserted into their sequences to trigger better targeting. Peptides can also be conjugated as linear or cyclic forms to improve stability and binding of AAV vector or non-viral vector (e.g. Lipid Nanoparticles, LNP). Optimization opportunities include incorporating non-canonical amino acids to improve pharmacological properties.

WhiteLab Genomics’ AI-powered platform pre-screened thousands of unique peptides extracted from our proprietary peptide library, employing a rational workflow: (i) Analysis of known binders to identify critical interaction motifs, (ii) Customization of computational tools tailored to receptor binding, (iii) Virtual screening of high-potential candidates, (iv) Structural optimization using SAR analysis, and (v) Peptides refinement through macrocyclization, sampling, and dynamic modeling.

Hendrik Goeddeke, AbbVie 

Applying FEP+ at large scale to score a reaction-enumerated library of two million ligands 

While free energy perturbation methods like FEP+ have become a powerful tool for accurate binding affinity calculations, the time and cost of running these calculations limits their broader application on (ultra-)large chemical spaces. Active-learning FEP (AL-FEP) combines short FEP calculations with machine learning (ML) to efficiently screen large compound libraries. First, a compound library was generated by reaction-based enumeration with AbbVie’s monomer library and filtering it down by PhysChem properties, PAINS/REOS and docking score resulted in a 2M on-target library. The initial set of 1500 molecules were chosen based on docking score and diversity for the first iteration of single-edge FEP. DeepAutoQSAR was then used to train the ML model based on the FEP scores of 1500 molecules. Since the output molecules turned out to be less diverse and have suboptimal properties, the acquisition function was adapted multiple times to project needs. Several compounds have been synthesized, and most are active against the target; a few of which show double digit nM activity and a favorable off-target profile. Furthermore, we have started to leverage the ML model generated by AL-FEP as a scoring component in generative AI approaches such as REINVENT4.

Matthias Bauer, Novartis 

Optimization of a mutant PI3Kα binding free energy protocol towards accurate potency prediction on a challenging target 

PI3Kα is a key oncology target, particularly in breast cancer, where it is often activated by mutation. Although FDA-approved PI3Kα inhibitors like PIQRAY® exist, there is a need for greater selectivity between mutant and wild-type forms for next-generation inhibitors. To address this, we have developed and optimized a binding free energy protocol to predict the potency of ligands that bind to the mutant-specific, allosteric PI3Kα H1047R pocket. Major challenges include the large size of the PI3Kα heterodimer, the cryptic nature of the allosteric binding site, and the high flexibility of the kinase domain activation loop.

Through collaboration with Schrödinger, the FEP+ binding free energy prediction performance was significantly improved by investigating various setup parameters and structural templates. The optimized setup has been extensively used in the project’s in-silico flowchart, which includes FEP+ potency predictions, Glide ensemble docking, conformational and property predictions at the quantum chemistry level, and various ML property models. Overall, the in-silico flowchart has greatly impacted successful compound prioritization towards more potent and LipE-efficient mutant PI3Kα inhibitors.

Shane Wald, QR Genetics  

Accelerating rare disease treatments: An AI-driven approach to genetic discovery  

QR Genetics leverages advanced AI-driven discovery tools to identify and address the genetic causes of rare diseases. Our integrated approach focuses on diseases that alter protein function, using ACTA2-related vascular disease as a key example. By combining proprietary AI technology with Schrödinger’s molecular dynamics simulations, we pinpointed the underlying genetic mechanism, validated the findings in cells and animal models, and identified a potential therapeutic candidate.

Our platform’s AI capabilities enabled us to prioritize an existing FDA-approved drug that was subsequently tested in human trials. The results were remarkable: the drug halted disease progression and even reversed life-threatening conditions. By seamlessly incorporating computational tools like molecular dynamics as part of our broader AI-driven process, QR Genetics transforms data into actionable therapeutic strategies.