Deep learning assisted, smartphone based universal Multi-RPA-CRISPR/Cas12a-G4 portable chip for simultaneous detection of CaMV35S and NOS

Abstract

Portable and fast genetically modified organisms (GMOs) detection can achieve accurate labeling and precise regulation. Here, we build a universal and intelligent Multi-RPA-CRISPR/Cas12a-G4 detection chip platform for simultaneous CaMV35S and NOS detection. The use of multiple primer pairs for RPA enables one-tube amplification of dual targets. Independent cleavage of Cas12a for dual targets was controlled by flow amplification products into the respective detection chambers containing CaMV35S crRNA and NOS crRNA. The detection results are obtained based that uncut G4 in solution can produce peroxidase activity to catalyze ABTS²⁻ making the negative result approaching dark green. The multi-RPA-CRISRP/Cas12a-G4 colorimetric platform achieves a 1 aM LOD for CaMV35S and NOS in 55 min. The Yolov5 deep learning algorithm and the analysis of gray values was trained and developed for colorimetric detection that is more sensitive than the naked eye. These algorithms are integrated into a smartphone app to automate the reading of test results. The 100% accuracy of colorimetry detection and the 93.3% accuracy of portable chip detection for rice and soy sauce samples demonstrate our method can rapidly and inexpensively detect samples of GMOs across species. Overall, our efficient, low-cost portable Multi-RPA-CRISPR/Cas12a-G4 detection platform coupled with deep learning algorithms enables portable GMOs detection with automated signal readout.