Pharos-3D

Overview
The Imperative for Advanced Virtual 3D Screening in ULCLs
A Customer Story
Pharos-3D: Iptacopan Case Study
Getting Started
Access and Pricing
References

Overview

Small molecules are a cornerstone of modern medicine, comprising the majority of approved therapeutics and showing a consistent rise in new approvals. In drug discovery, computationally assisted hit expansion via scaffold hopping and lead optimization are pivotal strategies for identifying structurally novel lead candidates that mimic shape and function of known active compounds. These approaches are essential for initiating new medicinal chemistry programs, as they provide pathways to secure and expand intellectual property (IP) and resolve liabilities related to ADMET (absorption, distribution, metabolism, excretion, and toxicity).

Pharos-3D is a novel computational 3D-similarity search method that leverages and combines 3D-shape and pharmacophore models for the efficient virtual screening of ultra-large combinatorial spaces (ULCLs). It integrates low-energy conformer shape matching with an analysis of potential protein interactions.

The Imperative for Advanced Virtual 3D Screening in ULCSs

1. Overcoming the Limitations of Experimental Screening
The pharmaceutical industry faces a critical demand for continuous access to diverse, safe, and efficacious hits. However, the traditional process of experimental high-throughput screening (HTS) has become a bottleneck due to its prohibitive costs and time-intensive nature. To maintain drug discovery momentum, there is an urgent need to shift from physical screening to computationally efficient methods that can deliver high-quality leads at a fraction of the cost.

2. Bridging the Software Gap for Ultra-Large Spaces
Current software solutions lack the capacity to efficiently access and explore ultra-large spaces containing trillions of "make-on-demand" compounds. To effectively mine these vast chemical spaces for bioactive compounds acting on specific targets, researchers require next-generation software capable of performing exhaustive searches across spaces from trusted vendors and internal databases without compromising on speed or depth.

3. The Necessity of Advanced 3D Algorithms
To improve the probability of finding viable drug candidates, virtual screening must go beyond simple metrics. The integration of innovative 3D algorithms is essential; these tools must connect shape similarity of the best-matching low-energy conformers to the query while simultaneously predicting potential protein interactions. This multidimensional approach ensures that the exploration of massive chemical spaces yields hits that are not only diverse but also structurally and energetically viable for further development.

A Customer Story

Pharos-3D: Iptacopan Case Study

As an expample result for Pharos-3D, below depicted are the best Pharos-3D alignments from each library with respect to the Pharos-3D Score together with the respective 2D projections of each compound. The Pharos-3D Score represents the shape similarity of the best matching low energy conformer to the query while considering potentially similar protein interactions. The score ranges from 0 to 1, indicating “no alignment” and “perfect alignment with equivalent interaction potential”, respectively.

For the six vendor spaces, we selected all Pharos-3D Scores larger than 0.7, and to retrieve structurally diverse hits from the result set, selected only hits with SkelSpheres Similarity smaller than 0.48. These selection rules yielded a set of 1395 hits, approximately 2.3% of all hits. To illustrate the similarity clustering of this selection, see below a similarity chart color-coded with respect to vendor library.

The SkelSphere Similarity measures 2D structural similarity based on the SkelSphere descriptor, and ranges from 0 to 1, indicating “no 2D structural similarity” and “2D structural identity”, respectively.

Getting Started

A virtual screen with Pharos-3D is performed via a plugin in DataWarrior by default, and via well-known commercial cheminformatics platforms by request. It is esssetnially the same than executing a Hyperspace search. Make sure, you have installed DataWarrior version v06.04.01 or newer. Then, install the Pharos-3D plugin, either by selecting Alipheron Pharos-3D from the Help->Trusted Plugins menu, or, if you have received a dedicated plugin for your own space, by putting that plugin-jar file into the DataWarrior plugin folder. Then, quit DataWarrior and re-launch it to get a new Alipheron->Pharos-3D Search... menu item. Select this item and the following dialog opens.

After selecting the space to be searched, you need to define your query conformer. A right mouse click in the query field opens a menu letting you paste in or load a molecule from a file or a ligand directly from the PDB database. You may change the protonation state of acidic or basic atoms and assign lower or higher importance to parts of the query conformer. Pressing OK submits your query definition to the server, which immediately starts to work on you request.

Pharos-3D searches are computationally demanding, because they involve the generation of conformers of thousands of partially assembled and also completely enumerated structures. A typical result of a Pharos-3D search contains thousands of molecules with 3-dimensional atom coordinates that match well the query structure. Typically, these molecules are ranked then by another method, e.g. ligand-protein docking, to determine a most promising subset to be ordered for synthesis at the provider of the space. For instance, docking can be performed using DataWarrior directly.

Access and Pricing

For licensing options, contact us at contact@alipheron.com.

References

• Flexophore, a new versatile 3D pharmacophore descriptor that considers molecular flexibility; M von Korff, J Freyss, T Sander; Journal of Chemical Information and Modeling, 2008, 48 (4), 797-810; https://doi.org/10.1021/ci700359j

• PheSA: An Open-Source Tool for Pharmacophore-Enhanced Shape Alignment; J Wahl; Journal of Chemical Information and Modeling, 2024, 64 (15), 5944-5953; https://doi.org/10.1021/acs.jcim.4c00516