Assay Classifier Engine (ACE) for enhancing splice sensor assay performance

Lucerna, Inc., Brooklyn NY

Investigators

Abstract

SUMMARY: The goal of this proposal is to improve the sensitivity and specificity of the Spinach-based splice sensor platform by developing a novel multiprobe (MP) assay design and a companion machine learning-based classification algorithm called assay classifier engine (ACE). Improvement in sensitivity and specificity of the splice sensor platform enables its application to detect endogenous RNA isoforms with low copy number and distinguish alternative RNA isoforms that share high degree of sequence similarities. The aim of any assay development effort is to achieve excellent assay specificity and sensitivity. However, this is often a futile endeavor since specificity and sensitivity are two inversely correlated factors. The underlying reason for poor sensitivity or specificity is due to the off-target signals generated by competing molecules present in the sample. In the field of diagnostics, one of the ways these issues are addressed is to perform multiple single probe testing instead of one single probe testing. While individual singe probe assays might have poor specificity and sensitivity, when combined, these assays synergistically improve the sensitivity and specificity of the ultimate diagnostic determination. In the field of research and drug discovery, researchers have employed a multitude of strategies (e.g. signal amplification, reaction cascades, or sample enrichment) to improve sensitivity and MP design or strand displacement strategies to improve specificity. Some of the PCR- based methods have combined both enzyme-based signal amplification and MP strategies to improve assay determination. However, when it comes to detecting targets that are highly similar to their competitors, such as detecting single nucleotide polymorphism, DNA methylation, RNA modification and alternative splicing, there is still an unmet need for more sensitive and specific analytical methods. In the past few years, Lucerna has developed Spinach-based sensors to detect intractable metabolites and biomolecules. One such sensor is the splice sensor, which is a Spinach-based sensor that can generate fluorescence signal based on the alternative RNA isoform of interest. One of the challenges encountered during splice sensor assay development is the lack of sensitivity toward low copy number RNA isoforms and low specificity when distinguishing two splice isoforms that share a high sequence similarity. To overcome this challenge in this proposal, we will develop a MP assay panel comprised of splice sensor variants that recognize the target RNA and the competitor with varying binding affinities and differing signal responses. We will use data sets generated from the MP assay to train a ML-based ACE algorithm to make target determination in test samples. Further, we will develop a quantitative MP data set and re-train the ACE algorithm to classify the assay signals into various categories based on target concentrations in the test sample. This new ACE algorithm will then be tested against conventional single probe assays to determine specificity and sensitivity improvement of the MP assay platform.