Data silos in discovery chemistry are still a problem
In today’s challenging small molecule discovery environment, innovation cycle time is more important than ever. Access to the right data at the right time and in the right place can help avoid costly excursions along unfavorable avenues of research. Poor access to data is recognized as being a significant blocker to efficiency in the pharmaceutical industry, with a recent survey of 400 industry professionals showing nearly half believed that data silos were derailing cross-functional collaboration in their organization.
Built on a deep understanding of these industry challenges and designed as a central location for data sharing and collaboration, Torx® Software has been created from the ground up to easily interface with a wide variety of data sources, systems, and services, breaking down silos and ensuring decisions are made in data rich environments.
Design on a single pane of glass
Torx Design–Analyze provides a user-configurable interface where plugins (dedicated windows that provide users with access to information or tasks) are combined to compare and contrast 2D and 3D molecular features on a single page. This enables users to create a single pane of glass containing the tools and data they need for the specific task they wish to perform.
Figure 1: 2D and 3D molecular features can be combined with data from external resources on a single page
Through this interface, Torx presents the user with data from external resources such as biological results from a data warehouse or CDMS (such as CDD Vault) or CSV and SDF files. Multiple sources can be merged, enabling the creation of Data Sets that contain both 2D and 3D representations of molecules, or enabling the merging of data that is stored in multiple locations. Designed molecules that have been through Torx and subsequently appear in connected sources are merged into a single row in the Data Table, enabling users to rapidly identify outcomes and inform the next round of design.
External information is not limited to data; any web-based content is accessible through Torx’s dedicated web component (WebFrame) that can access and post data to the current molecule enabling rapid deployment of internal methods and resources through the Torx environment. The currently selected design is sent to a server and the generated results (e.g. a list of design suggestions, calculation results, 2D or 3D representations, etc) returned and displayed to the user. Crucially, the generic WebFrame plugin enables a wide range of services to be developed and deployed independently, without the need for a specialized developer.
Use live feedback to inform design decisions
One scenario that will resonate with most chemists is where considerable time and effort has been expended in synthesizing a compound, only to find it already exists within a corporate repository (and may even be physically available).
To counter this risk, Torx connects seamlessly to corporate databases and provides the user with results of a similarity search as a molecule is sketched or selected highlighting exact matches but also giving nearest already existing neighbors. Where more extensive searching is required, a dedicated interface provides the ability to perform similarity and substructure searching in connected external services and databases directly from the Torx interface. Its seamless integration to CAS SciFindern also enables subscription holders to perform searches via a dedicated plugin window, enabling access to literature and patent reviews, chemical substances, reagents, and reaction schemes.
Figure 2: Editing a molecule updates results from a similarity search and recalculates physico-chemical properties from connected providers
As users create and edit molecules in Torx they can view in real-time how those changes might affect predicted physico-chemical properties. A dedicated ‘Properties’ plugin shows selected data for the current and any number of comparison molecules to enable users to make rapid meaningful decisions on molecule progression. Out of the box, Torx features an instantaneous RDKit property provider, but also enables connection to external property calculation or prediction providers, providing live results from custom models to the user as they design. This may include specific physico-chemical properties (e.g. clogP) or ADMET predictions from third-party providers, custom computational models developed on internal data, or activity data for existing compounds.
Integrate generative learning
User expectations of seamless access to data are higher than ever, with paradigm shifts evident in platforms for streaming and social media, where user preferences are learned and leveraged to make recommendations. Life sciences are following this trend with applied artificial intelligence and industrialized machine learning being two of the top three areas of tech investment for this sector.
Torx features a wide range of points through which AI-driven ideation capabilities and predictive models can be integrated to make provision of that information seamless to the end user. For example, Torx has an extensive API that enables external tools to query and modify molecules, post new suggestions, or update poses. Where more elaborate or custom interfaces are required for users to interact with outputs from (or make requests to) AI-based services, Torx’s WebFrame technology can be used.
Minimize context-switching during synthesis and testing
While seen as the norm in today’s modern work environment, the cost of users needing to toggle between multiple applications should not be underestimated. A recent study of 137 office workers across three Fortune 500 companies demonstrated that constant context switching between different applications and websites was costing each user around four hours per week, equating to approximately 9% of their annual time at work.
As a molecule moves through the design process and into synthesis, it accumulates metadata and supporting information which may be stored in multiple systems, leading to frequent context-switching. Furthermore, as synthesis is completed, this data must be extracted from those systems and input into the corporate repository for registration of the molecule. Since this task is often performed manually it not only occupies time that might otherwise be spent at the bench, but also incorporates a risk of input error, potentially leading to significant downstream repercussions. As a molecule is sent for testing, the same need to context-switch and perform manual entry once again applies: its identity and metadata such as structure, molecular weight, and purity must be input into a test request for submission to the relevant service provider. Across both Make and Test phases, the need to collaborate with external parties such as CROs adds a further layer of complexity, with fragmentation of data in emails and file shares making it difficult to keep track of work.
The Torx platform enables these issues to be avoided entirely, by acting as a central hub for information sharing across the DMTA cycle. Underlying the platform is a workflow that moves molecules in Design from idea status into a review process where decisions can be made on which to synthesize. Once agreed, molecules are transitioned into Torx Make’s chemically aware Kanban board for assignment to either internal or external chemists, and real-time synthesis tracking. Each of the Kanban tickets can hold a wealth of chemically relevant information, such as amount or chirality requested, detail on synthetic steps, or proof of purity such as LCMS data, and they can also serve as a jumping off point to relevant pages in the corporate ELN via integrated links.
Figure 3: Torx Make tickets contain a wealth of chemical information. Compounds can be submitted seamlessly to a connected corporate registration system on ticket status change
Torx Make can be seamlessly integrated with a corporate registration system, eliminating the need for repeated entry of data. Moving a molecule into the ‘Registered’ status on the Kanban board validates all mandatory information has been completed by the chemist and triggers submission of the molecule and its associated data to the registration system. Simultaneously, test requests to internal or external services are automatically submitted, and the status of those tests then tracked in Torx Test. Test service owners themselves can also be confident that assay names and descriptions match those defined in corporate systems, as these can be aligned through an integration point in Torx Test. In this way, context-switching across multiple systems and repeated manual entry of data can be avoided, and users can focus on what they do best – making the molecules that matter.
Summary
With information scattered across multiple data sources and systems, data fragmentation is still an important challenge in molecule discovery. As a central location for data sharing and collaboration, Torx has been designed from the ground up to easily interface with external data and services, enabling users to view information from across the entire DMTA cycle on a single pane of glass.
