HoloMoA: Fast holography and deep-learning-based tool for the novelty detection of mechanism of action of antimicrobial candidates

ABSTRACT

We propose an innovative technology to classify the Mechanism of Action (MoA) of antimicrobials and predict their novelty, called HoloMoA. HoloMoA is a rapid, robust, inexpensive, and versatile tool based on the combination of time-lapse Digital Inline Holographic Microscopy (DIHM) and Deep Learning (DL). In combination with proper image reconstruction, DIHM enables a label-free, time-resolved visualization of bacterial cell morphology and phase map (i.e. refractive index × thickness) to reveal phenotypic responses to antimicrobials, while DL techniques are powerful tools to extract discriminative features from image sequences and classify them. We assessed the performance of HoloMoA technology on Escherichia coli (E. coli) ATCC 25922 treated for up to 2 hours with 22 antibiotic molecules representing 5 functional classes (i.e. Cell Wall synthesis inhibitors, Cell Membrane inhibitors, Protein synthesis inhibitors, DNA and RNA synthesis inhibitors). First, using reconstructed phase images as input to a 3D Convolutional Neural Network classifier, we detected the MoA of known antibiotics with 89% accuracy. Secondly, we showed how our CNN models combined with a Siamese neural network architecture can be used for the novelty assessment of the MoA of a candidate antibiotic. The HoloMoA novelty detection tool succeeded in detecting trimethoprim-sulfamethoxazole (Folic Acid synthesis inhibitors) as belonging to a novel functional class (i.e. different from the 5 aforementioned classes). We demonstrated that combining DHIM and DL gives a promising tool for determining the MoA of new antimicrobial candidates provided that a large image database for known antimicrobials is available.