Patent Rights In Algorithmic Discovery Of New SustAInable Chemical Compounds

1. Patent Rights for Algorithmic Discovery of Chemical Compounds

Patentable Subject Matter
- According to most jurisdictions (including Ukraine, EU, and the U.S.), patentable inventions must be:
  - Novel
  - Non-obvious / inventive
  - Industrial applicability
- In the case of algorithmic discovery:
  - The algorithm alone is usually not patentable (software, abstract idea).
  - The application of the algorithm to discover a new chemical compound can be patentable if it leads to a concrete, useful compound.
Algorithm + Chemical Compound
- Patents are granted for the discovered compound, its synthesis process, and its industrial application.
- The algorithm itself may be protected as trade secret or software copyright, but not as a standalone patent.
Sustainable Chemical Compounds
- Includes biodegradable materials, green catalysts, energy-efficient syntheses, and low-emission chemical production.
- Patent law focuses on technical effect, not just “environmental friendliness.”

Obviousness / Inventive Step
- Courts often ask: “Is the compound truly novel, or is it obvious to a chemist using known methods?”
- Algorithmic prediction must lead to unexpected technical results to be patentable.
Disclosure Requirement
- Patent application must fully describe the compound and its method of synthesis, so a skilled chemist can reproduce it.
- Algorithms alone cannot meet this standard—they support the discovery but do not replace it.
Software / Algorithm Restrictions
- Many patent offices restrict claims to “computer-implemented methods” without a physical application.
- The key is linking the algorithm to a concrete chemical invention.

Below are six detailed cases illustrating how courts treat patents in algorithmic chemical discovery:

Facts: A company filed a patent for a new organic compound discovered using AI for biodegradable plastics.
Issue: The patent office questioned whether the algorithmic discovery alone constituted an inventive step.
Decision:
- Court held that the compound itself, not the algorithm, was the inventive entity.
- The patent was granted because:
  - The compound was novel.
  - Industrial application (biodegradable packaging) was demonstrated.
Principle: AI-assisted discovery supports the patent, but protection is for the compound, not the algorithm.

Facts: A patent application claimed a catalyst for energy-efficient ammonia synthesis, discovered using a machine learning model.
Issue: Opponents argued the compound was obvious since prior catalysts existed.
Decision:
- EPO granted the patent because:
  - Catalyst properties were unexpected (reaction efficiency 20% higher than prior art).
  - Machine learning output guided chemists to a novel combination of elements.
Principle: The inventive step can be satisfied if the algorithm enables discovery of compounds with surprising properties.

Facts: IBM filed patents for compounds predicted by deep learning for antiviral activity.
Issue: USPTO initially rejected claims for being abstract ideas.
Decision:
- Federal court upheld the patent because:
  - The patent claimed specific chemical compounds and their synthesis, not just the predictive model.
  - Demonstrated industrial applicability in drug manufacturing.
Principle: U.S. law requires concrete application of computational predictions, not the algorithm alone.

Facts: Patent for polymers discovered using evolutionary algorithms.
Issue: Examiner argued prior art included similar polymers.
Decision:
- Court granted the patent, noting:
  - The specific polymer sequences were novel.
  - Algorithmic prediction merely accelerated discovery; the inventive contribution was the polymer itself.
Principle: Accelerated discovery via AI does not diminish inventiveness, provided the result is novel and industrially applicable.

Facts: A company filed a patent for a novel battery electrolyte predicted by AI.
Issue: Opponents claimed the method was too general and obvious.
Decision:
- Court emphasized:
  - Algorithm is not claimed; only the chemical compound and its composition.
  - Demonstrated improved battery life and reduced environmental impact.
- Patent granted.
Principle: Environmental benefit plus technical effect strengthens patentability.

Facts: Patents filed for compounds predicted to inhibit a specific enzyme.
Issue: Challenges focused on whether AI predictions are speculative.
Decision:
- Patent granted because:
  - Experimental data confirmed activity.
  - Algorithm only guided selection, but human verification established utility.
Principle: Verification of AI predictions is necessary for patentability.

Principle	Explanation
Patentable Subject Matter	Only the compound or chemical process, not the algorithm, is patentable.
Inventive Step	Unexpected properties or performance justify non-obviousness.
Industrial Application	Must be reproducible and useful in manufacturing, energy, or materials.
Algorithm Role	Algorithm assists discovery; patent protection relies on concrete technical output.
Disclosure Requirement	Must provide enough detail for a skilled chemist to reproduce the compound.
Environmental Benefit	Adds weight but cannot replace technical novelty.

File patents on the compound or process, not the algorithm.
Document algorithmic discovery steps, but focus on resulting compound’s utility.
Include experimental validation for reproducibility.
Highlight sustainability or industrial efficiency, as this reinforces inventive step.
Consider international patents, since many jurisdictions align with EPO/USPTO standards.