Ultrasensitive bioanalysis and machine learning

While significant advances have been made in terms of ultra-sensitive technology platforms for measurement of biologics and biomarkers, most platforms requires high capital investment and significant re-work of existing immunoassays. We present case studies where we are able to adapt the current ELISA methods to achieve 5-50x more sensitivity. The adaptation involves utilization of nanoparticle enabled immune-PCR in combination with machine learning augmented design of experiment (DOE) to deliver robust immunoassay method.

Why Machine Learning DOE?

• Immunoassay methods are highly variable and impacted by numerous assay parameters and conditions.
• Current approach is to utilize trial and error approach – one factor at a time. Due to lack of understanding among variables/factors, when a reagent changes it is difficult to troubleshoot methods.
• It is difficult to describe these variables even with high degree of mathematical structures.
• Due to high number of variables impacting immunoassays, DOE by itself very time-consuming and expensive process.
• As the complexity of the methods increases, it requires expert knowledge to define numerous parameters and options and to analyze the data.
• Existing experimental design methods such as the classical techniques, Taguchi methods, and response surface methodology is not adequate to model the representation of many underlying multi-response ELISA problems and is not adequate to provide decision support.

Decision Support System

• Design of Experiments (DOE) is a statistical method used to determine the cause-and-effect relationships between different factors in a system. Machine learning, on the other hand, is a field of within Artificial Intelligence that focuses on developing algorithms and models that can learn and make predictions based on data.
• DOE can help identify the important factors that affect the performance of a system, and machine learning algorithms can then be used to develop models that predict the performance of the system based on those factors.

We have developed a Decision Support System (DSS) that utilizes DOE to identify the factors that affect the quality of the method , such as the temperature, antibody concentrations, and incubation time etc. Machine learning algorithms can then be trained on the data from the DOE to develop models that can predict the ideal conditions based on historical data and ongoing DOE data