How long does it take to implement the QuartzBio® platform?
Typically, we deploy infrastructure to new clients and deliver production data in the UI within weeks from contracting
Can the QuartzBio platform be deployed if a study is already in progress?
Yes, we routinely deploy during ongoing studies.
To what data specifications must our vendors adhere, so that data can be uploaded to the QuartzBio platform?
We’ve purpose-built the QuartzBio platform with the recognition and deep appreciation for the complexity and inconsistency of data. Our platform is source system- and format-agnostic. We do not enforce rigid data specifications that require vendors to change how they provide data.
With this in mind, we can also integrate data for studies that have already been completed.
How does data get into the platform?
The most common way that we engage with vendors is being added as a third party to an existing data transfer agreement between the vendor and the sponsor. We also frequently receive data uploads directly to the platform from sponsor teams.
We have the capability to integrate with APIs, though vendors don’t always have them available. With our breadth of experience, teams are often pleased to learn that QuartzBio is already working with their vendors.
QuartzBio is also flexible and open to aligning with existing processes and capabilities that sponsors have in place.
Can the QuartzBio platform function across multiple studies?
Yes. Not only can the platform integrate data across multiple studies, but it can also integrate historical data and publicly available datasets.
The key to this capability is QuartzBio’s standardized data model, which encompasses a variety of data elements covering subject and sample metadata as well as the variety of assay technology-specific metadata and reportables. However, the data model is flexible and can be amended to incorporate various additional data elements based on sponsor needs. This is usually supported by QuartzBio’s Professional Services team during the setup phase in collaboration with the sponsor.
Is this system GDPR compliant? How about 21 CFR Part 11?
Yes, read more about our key security features and compliant infrastructure.
How often is data refreshed in the platform?
Data is refreshed on an ongoing basis as results data becomes available. Frequency is largely dependent on the needs of the client/sponsor.
The QuartzBio platform can be used to identify and address discrepancies; do you have to wait until the clinical database is locked to do this?
No, it’s not necessary to wait until database lock. With the vSIM solution, data are integrated continuously throughout the life of a study. This approach is designed to be implemented before database lock, and discrepancies are highlighted as data is refreshed.
When updates are made in the source system, the discrepancies will be automatically resolved in the platform. With this approach, everyone on the sponsor and CRO teams can work from a common single source of truth – and they can confirm that source systems are updated appropriately. This dramatically reduces the burden of tracking discrepancies.
Our central lab does sample reconciliation – how does this system fit in with that process?
Reconciliation efforts typically have large gaps for sample information that is critical for sponsor teams. These gaps are often related to central labs being focused on reconciling samples they receive back to the requisition form or EDC. Gaps in this process include the ability to report in a timely manner on missed collections, unexpected samples, and the status of samples and derivatives as they move onto third-party testing labs.
In order to create a holistic view of samples on-study, QuartzBio pulls information together from the EDC, and each lab and biorepository. We also incorporate the sample schedule from the protocol, including changes to the protocol as new patients are enrolled and as the protocol is amended.
We’re running a trial with (name of specific CRO) using (name of specific EDC). Are these data sources compatible with QuartzBio?
The answer is, universally, yes. Our team has deployed this solution on hundreds of studies, and this solution is not limited to a specific list of vendors. We have experience with inventory data from most of the active CROs, EDCs, central labs, and specialty vendors – this even includes data from academic labs, which can be particularly challenging.
Central to our approach is that our team and solutions are flexible. If your trial involves a vendor we haven’t worked with before, it is straightforward to set up a data ingestion pipeline for a new data source. This is a process that takes weeks, not years, so it can be deployed even on active studies.
QuartzBio Virtual Assistant FAQs
What is QuartzBio’s Virtual Assistant?
- Our Virtual Assistant (VA) is our conversational AI capability, powered by a purpose-built Precision Medicine Large Language Model (LLM), that allows users to conversationally interrogate their sample, biomarker and clinical data housed within QuartzBio’s Biomarker Intelligence Platform. This allows activities that may take minutes or hours to be reduced to seconds.
What are the benefits of the Virtual Assistant?
- Make smarter decisions, faster by creating a business-friendly “data dialogue”.
- Proactively detect anomalies, uncover novel insights and act on trends with real-time metrics at your fingertips.
- Use natural language to empower teams with varying levels of data expertise
- Leverage platform to create reproducibility, scalability and consistency of insights at the speed of decisioning.
- Create metric-driven dashboards based on data dialogue to enable workflows such as study and vendor monitoring
Is QuartzBio’s Virtual Assistant secure? Does your Virtual Assistant maintain confidentiality?
- Our Virtual Assistant is a fully secure, closed implementation that maintains all existing elements of confidentiality and data security.
- Your data is only accessible by authorized users in your environment.
- The VA goes through the same Software Development Lifecycle including security testing and QA testing as any other product feature.
What guardrails are in place against hallucination (i.e., nonsense responses)?
- Guardrails are implemented to identify out-of-scope user queries, i.e., only valid user queries related to your stored data are processed into data queries and responses.
- Multi-agentic framework is used to ensure accuracy of responses by disambiguating user queries if needed.
- Full traceability is provided from user query to data query (i.e., user can see how their natural language query was translated by the Virtual Assistant).
How is the Virtual Assistant developed and validated to ensure accuracy?
- Integration of generative AI into QuartzBio products and the Virtual Assistant specifically follows our software development lifecycle (SDLC) processes, which includes validation of any front-end and back-end capabilities as well as validation of AI models
- Part of our implementation approach is to
- (1) build a foundational model for the domain specific use cases of our platform using a broad range of simulated sample and biomarker data (including training on industry and vendor terminology to understand the key metrics, statuses, and questions our clients use)
- (2) fine-tune the GAI model for a specific sponsor based on their data characteristics and
- (3) perform model validation by comparing answers received via the Virtual Assistant against the known truth. This validation is possible because the Virtual Assistant operates directly on the data captured in the underlying database (e.g., Master Sample Inventory).
Is Customer data used to train your LLM? If so, how is our data used to train your LLM?
- Customer data is not shared or used outside of your secure environment.
- Metadata describing your data asset is used to provide context to user queries when invoking the Virtual Assistant.
- All metadata describing your data asset is strictly segregated from any other Customer’s metadata.
What data can be queried by the Virtual Assistant?
Any sample, biomarker or clinical data that has been ingested and processed by QuartzBio into our structured database is accessible by the Virtual Assistant.
How do I learn more and gather more information about capabilities or pricing?
You can contact sales@precisionformedicine.com and we’ll walk you through the complete offering and pricing details.
