Potential Contribution to HORIZON-MISS-2026-05-SOIL-02

1. AI-Based Metagenomic Classification and AMR Gene Detection

EnICS Labs brings experimentally validated expertise in AI-driven metagenomic analysis, focusing on bacterial classification and pathogen detection in complex environmental DNA datasets. Our group has developed deep learning pipelines combining convolutional, transformer, recurrent and state space models for taxonomic classification of microbial communities from shotgun sequencing reads. These methods are designed to handle sparse, low-coverage data, absent reference genomes for many soil-borne species, and ambiguity introduced by horizontal gene transfer.

Within this call, EnICS will adapt its platform for detection and classification of antimicrobial resistant genes (ARGs), mobile genetic elements, and antibiotic biosynthesis gene clusters in soil metagenomes. Our self-supervised learning models enable taxonomy-free classification of novel species, which is critical for the large fraction of soil reads unassigned in existing databases. Relevant published works include GCOC (genome classifier-on-chip, IEEE TBioCAS 2024), DASH-CAM (approximate search for genome classification, IEEE/ACM MICRO 2023), ViTAL (transformer-based lineage assignment, Bioinformatics 2024), and DIPER (pathogen detection, IEEE TC 2023). EnICS also participates in the EIC Pathfinder project BioPIM (grant 101047160), developing a European bioinformatics processor.

2. Portable Field Hardware Platform for Soil Analysis (TRL 6–7)

EnICS possesses proven capability in complete hardware development, from concept through VLSI design, silicon fabrication and post-silicon validation, to high-TRL prototypes. Our lab has taped out over 40 ICs in advanced CMOS nodes (TSMC 16nm, 7nm) and validated multiple SoC platforms integrating custom AI accelerators with general-purpose processors.

For this project, EnICS can design and prototype a portable, field-deployable platform for real-time metagenomic analysis of soil samples. The platform integrates a dedicated AI inference engine based on our proprietary architecture with an embedded sequencing data interface. This enables energy-efficient, on-site classification of ARGs and microbial taxa directly from raw reads, eliminating the need for cloud connectivity in the field. The hardware will interface with portable sequencers (e.g., Oxford Nanopore MinION), providing end-to-end sample-to-result capability in a compact, battery-powered form factor suitable for deployment at contaminated or remote soil monitoring sites.

Leonid Yavits, Alexander Fish – EnICS Labs, Faculty of Engineering, Bar-Ilan University, Israel