Computational methods for discovering novel antifungal small molecules active against plant pathogens.

Bioenergy feedstock is valued at $116.5 billion (as of 2021), and is expected to reach $229 billion by 2030, registering a CAGR of 7.7%. Fungal pathogens are among the costliest diseases of bioenergy feedstock, with impact on corn alone exceeding 35 billion dollars annually. Concerns over the potential impact of synthetic fungicides on the environment, and the enaction of the Food Quality Protection Act in the United States in 1996 has brought natural products, as a more environmentally friendly alternative to their synthetic counterparts, to the center of attention for fungicide development. The classical natural product-based drug discovery starts with phenotype screening of extracts of natural samples to identify a bioactive 'hit', followed by iterative bioactivity-guided fractionations. However, this process is deemed to be insufficient since it is time/labor-intensive and often leads to the rediscovery of known molecules as it focuses mainly on the most abundant bioactive compounds from the cultivable microbes. The overarching goal of this proposal is to develop a platform for scalable discovery of novel fungicide natural products from plant microbial isolates using a computational analysis of low-cost genomics / metagenomics and tandem mass spectral data collected from the microbial isolates / communities. The focus of this proposal is on non-ribosomal peptides (NRPs), an industrially important class of natural products with applications in crop-protection and medicine. Phase I of this proposal delivers NP-Discover, a software for large-scale discovery of antifungal NRPs through the following steps. In the first step accurate machine learning methods are developed for predicting the chemical structure of NRPs from their biosynthetic pathways in microbial genomes. In the second step, machine learning methods are developed to detect and prioritize NRP biosynthetic pathways with antifungal activity. Since methods used in the first step might miss novel amino acids and modifications, in the third step predicted chemical structures are searched against mass spectral data in modification tolerant mode, in order to recover missing amino acids / modifications. NP-Discover will be tested on proprietary and public microbial genomics and mass spectral datasets, and one of the identified molecules with predicted activity against fungal pathogens of bioenergy feedstock will be further selected for HPLC-based purification, NMR and in-vitro testing. NP-Discover will be delivered by February of 2026, and it will be used for the internal discovery program. Revenue will be generated by out-licensing bioactive small molecules to agrochemical industry in deals that include upfront, milestone and royalty payments. An alternative strategy for generating revenue in the next few years is by providing NP-Discover as a software-as-service to the agrochemical companies that own large collection of micro-organisms.