Making data useful

David Hardy reveals how LIMS and data analytics deliver actionable insights

The biopharmaceutical lab is rich with data that helps drive decisions on which projects to progress. But although most think of data as being obtained from experiments, every single process in the lab provides an untapped goldmine of data waiting to be utilised – from quantities of reagents remaining to instrument health. All this data, when used well, offers many advantages to the lab, including: better uphold of data integrity; enabling compliance with regulations; improving productivity.

But collecting and using this data isn't as easy as one might initially think. Many laboratory processes are still manual, meaning the data is not automatically collected. And in cases where the data is collected, often it is siloed – meaning that scientists are not able to connect the dots to gather real-time updates.

Insufficient quantities of data, and data that is unconnected, can slow down laboratory processes, stifling productivity and innovation. Not only does it mean that the data cannot be used for analytics, but it also limits collaboration and, therefore, project progression. Decisions are also still being made after the fact and not in real-time, causing delays and incurring costs.

Laboratory information management systems (LIMS) are the missing puzzle piece that bring together data from different sources across the lab and enable it to be turned into information that allows users to take immediate action. Here, we illustrate how LIMS can be used for lab data management, allowing data to be leveraged to increase efficiencies and cut costs.

LIMS gather information from many processes digitally across the lab, such as the dates samples are sampled, received, or completed. Logging this data for every sample provides a wealth of information that can be used to make the lab more efficient.

Obtained data can be used in many ways – like for data mining – allowing the user to gain additional insight on lab performance indicators such as the number of samples run over a given time frame. Furthermore, users can easily identify bottlenecks, allowing them to fix issues, and, ultimately, reduce costs.

The LIMS data goes well beyond samples and can include aspects from training records through to unstructured data recorded in electronic analyst notebooks. The collected data can be broadly grouped into three distinct categories: lab operations, system administration and scientific insights. These combined areas lead to a highly detailed data set ready for further analytics.

Data analytics solutions can turn data collected by LIMS into actions. And to simplify matters, some LIMS have data analytics integrated – meaning that there is no need to export data, and data governance is maintained. Automated data analytics can assist with two main areas: business intelligence (BI) and machine learning (ML).

The ability to rapidly identify operational and administrative bottlenecks is essential to the smooth running of the lab. BI dashboards are tools that enable lab managers to gain a deeper understanding and turn data insights into actions. Depending on the LIMS used, different business intelligence capabilities will be available.

One of the crucial aspects of the biopharmaceutical lab, for instance, is ensuring a sufficient supply of reagents and consumables. Running out of these vital components can cause significant delays, forcing labs to miss project deadlines. Some LIMS come with a stock overview dashboard (Fig. 1), allowing users to view any consumable's location and availability – and even order new stocks. This information leads to a better understanding of consumables usage and distribution to help manage lab workloads.

Stock overview is just one example, though – dashboards cover many other aspects critical to the business, including instrument uptime, analyst workload and instrument maintenance.

ML can be applied to LIMS data in several ways to provide the biopharmaceutical lab with crucial scientific insight, including making predictions.

Training an ML model using high quality historic data can enable the prediction of future result values. Insights such as this provide a foundation to prioritise projects as the results data is still being obtained.

Progressing projects that are less likely to fail saves money. But what if this could be done with fewer tests? Labs can apply ML to LIMS data to find the relative importance of each test on the overall outcome. This allows labs to run the most important tests first, subsequently progressing the candidates with the highest promise.

In one test study, ML was used to predict drug activity. Investigating a dataset of 1,700 small molecules, the research team looked at 32 different chemical properties (Fig. 2). Training the ML model found the variables that are most important to the compound's activity: LogP, number of rotatable bonds, polar surface, and molecular weight – all of which, unbeknownst to the system, are key Lipinski descriptors.

Data is absolutely vital to the success of a lab, but only when it is collected properly and connected digitally. LIMS and their associated data analytics solutions effectively gather data and enable actionable insights to support lab operations, system administration and scientific insight. Only by embracing these capabilities can labs across the world drive decision making for increased efficiency, lower costs and ultimately, develop and deliver effective medicines more rapidly.

David Hardy, PhD, is senior manager, Data Analytics and AI Enablement at Thermo Fisher Scientific