Evogene (EVGN) AI Fungicide Discovery Yields 15 Active Molecules

On its second-quarter earnings call on August 18, 2026, Evogene (EVGN) said its AgPlenus unit is nearing the completion of lead optimization in the AI fungicide discovery program it runs against Septoria, and that the molecules it has synthesized are now in advanced biological assays, with greenhouse and field trials not yet started.[1]

Where ChemPass AI sits in Evogene's business

Evogene Ltd. (EVGN) is an Israeli computational biology company that neither manufactures nor sells finished crop-protection products; its revenue comes mainly from collaboration and licensing agreements and from a castor seed business. The crop-protection work sits in its wholly owned subsidiary AgPlenus, which uses a tool called ChemPass AI.

AgPlenus researchers working on herbicides and fungicides first ask the tool to predict which proteins inside a weed or a pathogen can serve as targets. They then ask it to pick molecules likely to work from existing compound libraries, and more recently to generate entirely new molecules directly. Chemists synthesize those molecules, and the laboratory tests whether they inhibit the target protein and whether they kill the pathogen. Only the ones that pass move into optimization, and only after that come greenhouse and field trials. The tool's job is to filter for molecules worth the investment before expensive synthesis and testing begin, which places it in AgPlenus's core business rather than in a support function.

Three years of disclosure, from posture to numbers

On the November 15, 2023 call, management could only say that outside interest and validation results were both improving, while acknowledging the technology was still early stage and needed more proof; there was no checkable output.[2]

On March 7, 2024, the herbicide collaborations with Bayer and Corteva connected that output to an external licensing channel: the AI designs and optimizes molecules against a newly discovered mode of action, and the partner pays an upfront payment, research funding and milestones.[3]

On November 21, 2024, checkable numbers appeared for the first time. Three AI-predicted proteins were experimentally confirmed essential to the pathogen's survival, and roughly 1,000 small molecules predicted to be effective against them went into in-vitro testing.[4]

On May 20, 2026, the tool moved from picking existing molecules to generating new ones based on the successes and failures of the preceding rounds.[5] By August 18, 2026, that same Septoria program was nearing the completion of lead optimization, and the company had added a model that predicts a molecule's activity inside the pathogen itself.[1]

The three screening rounds are the hardest evidence

In the first round, ChemPass AI selected 440 candidate molecules from compound libraries for testing. Only 11 met the enzymatic inhibition threshold, and 2 of those showed antifungal activity. In the second round, those 2 compounds, validated both in vitro and in vivo, were fed back into the system together with the negative results, and active search selected 164 off-the-shelf molecules for purchase; five of them showed antifungal activity. In the third round, a generative model produced 27 entirely new molecules that were sent for custom synthesis, of which 25 met the enzymatic inhibition threshold and 15 also reached the desired biological activity, which management called a dramatic improvement.[5]

Two limits sit on those figures. The pass standards behind the two thresholds have not been publicly defined, and the three rounds differ in where the molecules came from and how many there were, so the change in hit ratio cannot be converted directly into an improvement in success rate. The company has also published no before-and-after comparison of development cycle time or spending.

How laboratory hits would become revenue

AgPlenus licenses the molecules it discovers and optimizes to large agrochemical companies, which take on later development and commercialization, paying milestones as a program advances and royalties after launch. The number of laboratory hits therefore first determines how many programs reach a stage where licensing can be discussed, and only then, through a new collaboration or a milestone under an existing one, does it reach collaboration and licensing revenue.

As of today that chain holds only as a direction. Every public number stops in the laboratory, no payment or revenue has been attributed to this tool, and no amounts have been disclosed for the two herbicide collaborations.

What is confirmed and what is not

Confirmed: this fungicide program has been pushed to the end of lead optimization, the company holds a batch of molecules that passed laboratory testing, and it therefore has something to take into licensing discussions. Not confirmed: how those molecules perform in the greenhouse and in the field. The event that would connect the count of laboratory hits to revenue is a licensing or milestone payment carrying a specific amount.

Application assessment

• AI Crop Protection Molecule Discovery | Business position: Core business | Application stage: Pilot | Deployment scope: Single business unit | Value type: Revenue growth

Sources

[1] Drillr · Evogene Ltd. (EVGN) · August 18, 2026 · Earnings call

We have made substantial progress in our program to develop a novel fungicide targeting Septoria. We are nearing the completion of step 3 lead optimization of the campus AI process and we are currently testing synthesized molecules in advanced biological assays ahead of launching greenhouse and field trials.

[2] Drillr · Evogene Ltd. (EVGN) · November 15, 2023 · Earnings call

Yes, it's still in early stage. Yes, we still need to demonstrate more and more the power of our technology, but the company is definitely in the right direction in achieving this target.

[3] Drillr · Evogene Ltd. (EVGN) · March 7, 2024 · Earnings call

AgPlenus announced signing of a licensing and collaboration agreement with Bayer's Crop Science division. Under the agreement, AgPlenus will use its AI-driven computational modeling technology to design and optimize the molecules identified for their broad-spectrum herbicidal activity, targeting the APTH1 protein, a new mode of action identified by AgPlenus.

[4] Drillr · Evogene Ltd. (EVGN) · November 21, 2024 · Earnings call

The second milestone using ChemPass AI, the company identified approximately 1,000 small molecule compound, predicted to be effective against these three protein targets.

[5] Drillr · Evogene Ltd. (EVGN) · May 20, 2026 · Earnings call

Utilizing pointed the initial phase of molecule screening employing ChemPass AI, 440 candidates were selected for testing.

From these 2 in vitro and in vivo validated compounds and incorporating the negative results, utilizing active search the subsequent phase of molecule screening that leverage the data generated in the preceding stage, we selected 164 off-the-shelf molecules for purchase.

Building on these insights, we moved to lead up GPT and generated 27 novel compounds, which were custom synthesized and tested over the past months. Of these 25 met enzymatic inhibition thresholds and 15 also showed the desired biological activity, which represents a dramatic improvement.