How to reduce human intervention in analytical workflows without losing scientific control

A phrase like ‘less human intervention’ may sound uncomfortable to many working in an analytical labs because it’s easy to misinterpret as ‘less human oversight’. But in a well-designed analytical workflow, reducing human intervention is about removing routine handling from steps that can be better executed by the system while maintaining human oversight.

Keeping scientists in the loop is always essential since every analytical workflow extends well beyond the measurement itself.

There’s no value if we simply increase automation at one point in the workflow sequence while the scientists then have to spend additional time reorganising or manually interpreting the resulting data.

At CLADE, we look at bioanalytics from the workflow perspective. This is why we integrate sample analytics on the MIRA Analyzer with chemometric data processing in Sphere. We want to actively reduce operational effort between the sample and the final result without ever removing the scientist’s control and ability to examine the underlying analytical information.

Where can you reasonably reduce human intervention in analytical workflows?

There are certain parts in every analytical workflow that need attention but perhaps not continual scientific judgement, and these parts are ideal candidates for automation.

The MIRA Analyzer, for example, features automation on three levels:

1. The autosampler reduces sample handling time, while custom measurement jobs and job queues reduce the need for repeated manual loading.
2. Automated maintenance routines and autonomous cryocooling keep the instrument ready between runs, so the remaining manual steps are the short daily checks rather than continuous supervision.
3. Semi-automated data analytics speed up time-to-result thanks to chemometric modelling suited to the analytical question.

This means your scientists spend less time repeatedly advancing the instrument or data workflow while retaining access to the spectral information that goes into subsequent analysis.

“The purpose of automation is to remove routine intervention from the analytical workflow without removing the scientist from the decisions that depend on the data.”

Dominik Sievert, CEO at CLADE

How to apply automation without making data inaccessible?

When it comes to measuring multiple analytical attributes in biologics formulations with MIRA Analyzer (based on FTIR spectroscopy), spectra typically come with overlapping signals. This is why substantial data processing is required to access useful analytical results.

Sphere, the data analytics software that works with MIRA, is designed to reduce that processing burden so the system can handle data processing itself and check results before giving the user the output.

An example of retained data accessibility after automation is the data analytics tool that runs in Sphere, QA Scan. This tool has been designed for biopharma teams who need fast, standardised reports on multiple quality attributes in their formulations.

For a defined biologics formulation, QA Scan uses chemometric modelling and the SpecHub^TM spectral database to transform MIRA spectra into predicted values for selected quality attributes, such as:

• Protein concentration
• Excipient and surfactant concentrations
• pH
• Protein identity based on spectral similarity
• Protein secondary-structure signals

In instances where method and acceptance criteria have been defined, QA Scan can convert the measured spectra into endpoint values that support QC decision-making alongside existing specifications and validation package (with the scope of use defined per implementation). The real gain here is that routine spectral processing happens with minimal manual effort by the operator.

For situations where individual samples require more of a scientist’s attention, you can use the software application CLADE Workbench. It allows scientists to visually inspect FTIR spectra before QA Scan analysis to investigate anomalous samples or other scenarios that require case-by-case review. That creates an explicit place in the workflow for scientific review: the software can get on with routine processing while analysts can inspect spectral quality or identify outliers and then decide whether the data are suitable for subsequent analysis.

How to balance automation with human control for higher operational efficiency?

Returning to our opening point around less human intervention, you hopefully now see how this comes down to balancing operations that can be specified in advance and decisions that depend on scientific context.

Automated sample queues, instrument maintenance, preprocessing and defined chemometric calculations can all help reduce repeated manual work because the required operations are already specified.

Quality checks and interpretation, on the other hand, need a different relationship with the software; one where the system can organise data, calculate endpoint values, apply configured methods and provide confidence indicators, while always giving the scientist access to the spectra and analytical tools they need to investigate any unexpected results.

If your scientists are repeatedly loading samples or performing standardised data processing, automation can be a huge asset because it reduces operational effort and handling variability. Where the task is about assessing data quality, investigating an outlier or deciding how an analytical result should be used, evaluation by your scientific staff is always needed and they should have access to all the information they need to make an informed decision.

To learn more about increasing analytical workflow efficiency and end-to-end throughput, read this blog article next: How to achieve high throughput in biopharma analytical workflows