With the status quo, an analysis process could easily take up 2 months. Here’s an example of a scenario:

There’s a researcher, or an executive, who wants to ask a question, so they approach a data scientist for help. The data scientist then requests help from data engineers to find the correct datasets and piece them together. After that work is done, the data scientist then performs the ad-hoc analyses to answer the original question. Finally, all this work results in one-off projects that cannot be reused in the future.

Now, if you productize your data, you can dramatically reduce the analysis turnaround time – from 2 months to 2 minutes.

This is possible because Tag.bio data products are built to be useful – enabling researchers to instantly ask and answer their own questions, and enabling data scientists to integrate controlled, high quality data into their algorithms. The data products are owned and built by cross-functional data teams which ensures the quality and reliability of the product. With this approach, researchers can get answers to complex questions in minutes and data scientists can support multiple analysis projects at the same time.

Each data product is owned by a cross-functional data team. This approach ensures the quality and reliability of the data products. The team typically includes three roles: data engineer, data scientist, and researcher (i.e. domain expert).

The data engineers map and model data within each data product to make the process of data access and app building streamlined and flexible for data scientists.

Let’s go through a simple scenario: You want to standardize the use of the concept “Immune Subtype”. There’s a file with a table named “mmc2.txt.gz” that represents a table of data. Within this table, there’s a column named “Immune Subtype” with values of “C1, C2, C3” and so on. But many people don’t always know what those values mean. With Tag.bio’s data product mapping process, however, the data engineers can standardize the process and help other users the following ways:

• For data scientists, they’ll get an easy way to reference the column as “Immune Subtype” instead of constantly typing “select [Immune Subtype] from mmc2“. Additionally, they’ll be able to use the readable reference name everywhere as they build their analysis apps
• For researchers, they’ll get a human-readable version, such as “C2 – Interferon Gamma” instead of merely “C2“

And when there’s an updated version of the data, the data engineer doesn’t need to re-map the data because they had already set a good mapping foundation from the beginning.

Using Tag.bio’s SDK, data scientists and bioinformaticians can build point-and-click analysis apps for data products, integrating their R, Python, ML/AI code as business logic for algorithms and visualization.

For example, one data scientist can build an app which performs unsupervised clustering and another data scientist can build an app producing an R markdown report using the same data that has been mapped into the data product. Even though these are two different apps, they still use the same data model and controlled vocabularies thanks to the robust data mapping.

This approach enables the data scientists to build apps quickly without wasting time repetitively accessing, loading, cleaning, and modeling data. Additionally, researchers can use different apps to quickly analyze and cross-compare their findings because all the apps within a data product share consistent data and terminologies.

The researchers within a data team help the data engineers with the data mapping process by aiding them in identifying and standardizing data vocabularies that will be understood by other researchers. They also assist the data scientists in designing apps to answer a wide range of business and clinical questions.

The researchers outside the data team would use the no-code, self-guided analysis apps within the data products to ask and answer their own questions. For example, a researcher can perform an unsupervised clustering analysis with confidence because the app was built with guidance from other researchers who are either data owners and/or subject matter experts.

For researchers

• Ask and answer your research questions with confidence – by using self-guided, no-code analysis apps within each data products to perform advanced analytics, such as clustering analysis and generating R reports
• Promptly identify, store, and share your useful data artifacts (UDATs), such as biomarker and target candidates, with your fellow researchers and collaborators

For data scientists and bioinformaticians

• Demonstrate the value of your data science in clinical research – by making your work accessible to researchers and other domain experts across the organization
• Save hours from performing one-off analyses – by repurposing your work as reusable analysis apps across multiple data products

For leadership

• Promote a consistent view of reliable, harmonized research data – by providing a platform that presents a single source of truth
• Boost efficiency and drive innovation – by streamlining the data analysis process and making data science accessible across your organization

For strategic IT

• Simplify your data governance process as your research data grows – by streamlining your data management and access
• Stay compliant with regulatory and legal requirements without compromising data use

For data scientists

• Access reliable and harmonized research data from data products for your R & Python algorithms and ML/AI models
• Collaborate more efficiently when your fellow data scientists and data engineers use the same tool to access data and build data products

For data engineers

• Deliver quality and harmonized research data to data scientists and promote data reusability – by standardizing how nomenclatures get mapped into the data products
• Advance data literacy when you use terminologies that researchers understand – by partnering with domain experts to standardize the terminologies being mapped into the data products

For quality controlled analysts

• Rapidly reproduce any reports and analyses to support regulatory submissions and publications – with instant access to all the versioned data, software packages, and analysis methods
• Streamline ​​the auditing procedures when you gain transparency into the data mapping process and source code

For leadership

• Set a lasting foundation that promotes consistent quality controlled outputs as your research data grow and new data types emerge
• Significantly free up your resources time so that they can focus on what they do best – by streamlining and automating manual repetitive work