Ai-driven de novo antifungal peptide design platform for crop protection
RevivBio combines deep learning generative models with ultrahigh-throughput microfluidic screening to design and optimize novel antifungal peptides. The platform enables rapid discovery of potent, selective fungicides and host-plant incorporated peptides to combat fungal pathogens and manage resistance.
Overview
RevivBio is developing a platform that integrates generative deep learning with ultrahigh-throughput (uHTP) experimentation to design and optimize antifungal peptides (AFPs) for use as fungicides and in host-plant protection. Natural AFPs represent only a tiny fraction of possible peptide sequence space, leaving vast unexplored territory for discovering new candidates. By combining AI-driven sequence design with microfluidic functional screening, RevivBio aims to rapidly build a pipeline of potent, selective, and safe antifungal compounds that address the shortcomings of natural peptides, including toxicity to non-target organisms, low stability, and low production yield.
Technical specifications
The platform is built on two complementary technologies:
- Generative deep learning models: A pre-trained variational autoencoder (VAE) is fine-tuned via transfer learning on a curated dataset of 1,836 antifungal peptides. A regressor trained on reported Minimum Inhibitory Concentration (MIC) values against Botrytis cinerea predicts potency for newly designed sequences, enabling intelligent exploration of sequence space.
- Ultrahigh-throughput microfluidic screening: Synthetic biology and droplet-based microfluidic technology enable cell-free synthesis and encapsulation of individual peptides or peptide pools in picoliter-sized droplets. Each droplet serves as an independent assay compartment, supporting massively parallel functional testing at rates exceeding 1,000 tests per second and screening more than 10^12 protein variants per day.
- Design-build-test-learn cycle: Streamlined iterative cycles allow experimental data to continuously refine the AI models, optimizing both potency and selectivity. Designed peptides are tested against target fungi such as Fusarium species and counter-tested on non-target mammalian cells to ensure safety profiles.
Technology readiness level
The core uHTP protein engineering platform has been proven in antibody discovery applications. For antifungal peptide development, the team has assembled the necessary computational components, including the pre-trained VAE and MIC regressor, and is now applying transfer learning to the AFP-specific dataset. The next stage involves experimental validation of de novo designed peptides on the microfluidic platform, with iterative AI-guided optimization to advance candidates toward development-ready fungicide formulations or transgenic plant applications.
About RevivBio
RevivBio is a biotechnology company that creates next-generation biologics and enzymes by integrating a high-throughput microfluidic experimental platform with advanced AI and machine learning models. The company uses generative AI, quantum mechanics, and molecular dynamics to explore sequence space for protein design, allowing for the hyper-speed development of functional biologics and performant enzymes across markets including human health, crop protection, and animal health. Its technology platform enables the rapid screening of large gene libraries to identify variants with enhanced catalytic activity, stability, and substrate specificity.
The company is currently focused on developing enzymatic bioremediation solutions to address environmental contamination, specifically targeting the degradation of PFAS. Backed by the National Science Foundation, this work aims to provide a scalable, cost-effective, and environmentally sustainable alternative to current energy-intensive remediation methods. RevivBio has participated in programs such as Intel Ignite and NYU’s Endless Frontier Labs, and it actively seeks strategic partnerships to apply its bio-inspired design capabilities to diverse industrial and environmental challenges.
delivered to your inbox.