A stepwise guide to navigating multiplex immunoassay data analysis

A robust and accurate data analysis and interpretation pipeline is critical for gaining actionable and meaningful insights from multiplex immunoassays, as well as for ultimately identifying and validating protein signatures for clinical advancements. As multiplex protein analysis can generate large datasets, the management, quality control, analysis and interpretation of the data requires a uniform process, optimized with software and a set of tools that can ensure high data consistency and a robust data analysis process.

Here are four key steps to consider in the statistical pipeline of multiplex immunoassays:

1. Data acquisition and clean up: Getting the data together

Firstly, the data is combined with its clinical/experimental annotations (ex. cases vs controls, severe vs mild infection, etc.). In the case of multiple datasets, a normalization step is required for bridging the projects together, in which medians from overlapping/bridging samples are adjusted, thus making the datasets comparable. Finally, the data is cleaned up by addressing potential missing titles, mis-formatted datasets, QC warnings, and removal of samples that were found to be non-representative in later steps of analysis.

2. Quality control and exploratory data analysis: Does the data look as expected?

Next, the overall shape of the data and specific data points are assessed, with the aim of determining whether the data should be used as is, or whether editing is needed. Some of the most common ways to do this include using principal component analysis plots and looking for outliers/odd, non-normally distributed data, other unusual configurations, or samples that are not within standard ranges.

3. Statistical analysis: Using the data to answer your biological question

After the acquisition, clean up and QC steps, the data should correspond to the researcher’s basic assumptions. A statistical test that matches the study and that is most fit to answer the biological question should then be selected.

4. Biological interpretation: How are these proteins related to your biological question?

Once the data analysis is finalized, further research insights can be obtained by applying additional biological context, for instance, by assessing how the differentially expressed proteins are related to the disease or pathway of interest. This process can be guided by using a comprehensive database of annotations, which provides data- and literature-derived information on pathway coverage and disease-related biomarkers, with scores and rankings of connections.