Patent Eligibility Of Hybrid AI-Bioinformatics Technologies In Public Health Research
I. Legal Framework: Patent Eligibility (Section 101)
Under 35 U.S.C. §101, inventions are patentable if they fall within:
- process
- machine
- manufacture
- composition of matter
However, courts recognize judicial exceptions:
- Laws of nature
- Natural phenomena
- Abstract ideas
The Alice–Mayo Two-Step Test
This is the controlling test today:
Step 1: Is the claim directed to an ineligible concept?
Step 2: Does it add “significantly more” (an inventive concept)?
👉 This framework is critical for AI-bioinformatics systems, because:
- biological data → often “natural law”
- AI algorithms → often “abstract ideas”
II. Key Case Laws (Detailed Analysis)
1. Mayo Collaborative Services v. Prometheus Laboratories (2012)
Facts
- Patent claimed a method to optimize drug dosage using metabolite levels.
Issue
Whether a diagnostic method based on natural correlation is patentable.
Judgment
- NOT patentable.
Reasoning
- The correlation between metabolite levels and drug efficacy = law of nature.
- Additional steps were “routine and conventional.”
Principle
- Merely applying natural law using routine steps is insufficient.
Relevance to AI-Bioinformatics
- AI models predicting disease risk from biomarkers face the same issue:
- If AI only applies known correlations → not patent eligible.
👉 This case restricts diagnostic AI patents heavily.
2. Association for Molecular Pathology v. Myriad Genetics (2013)
Facts
- Myriad patented BRCA1/BRCA2 gene sequences linked to cancer.
Issue
Are isolated genes patentable?
Judgment
- Natural DNA → NOT patentable
- cDNA (synthetic) → patentable
Reasoning
- DNA exists in nature → discovery, not invention.
Principle
- Products of nature are not patentable, even if isolated.
Relevance to AI-Bioinformatics
- Raw genomic datasets used in AI:
- Not patentable themselves
- BUT:
- engineered datasets, synthetic constructs, or modified sequences may be patentable
👉 Important for genomic AI pipelines.
3. Alice Corp. v. CLS Bank International (2014)
Facts
- Patent on computerized financial transaction system.
Issue
Is implementing an abstract idea on a computer patentable?
Judgment
- NOT patentable.
Reasoning
- The claim was an abstract idea (intermediated settlement)
- Generic computer implementation ≠ inventive concept
Principle
- “Computerizing” an abstract idea is not enough.
Relevance to AI-Bioinformatics
- AI = mathematical models → often abstract ideas
- If claims only say:
- “Apply machine learning to health data” → NOT patentable
👉 Requires technical improvement, not just application.
4. Bilski v. Kappos (2010)
Facts
- Patent for hedging risk in energy markets.
Issue
Is business method patentable?
Judgment
- NOT patentable.
Reasoning
- Introduced machine-or-transformation test (not exclusive).
Principle
- Abstract ideas cannot be patented even if useful.
Relevance
- Early foundation for rejecting:
- AI models as mere “mental processes”
- Data analytics in public health
👉 Important precursor to Alice.
5. Ariosa Diagnostics v. Sequenom (2015)
Facts
- Method detecting fetal DNA in maternal blood (non-invasive test).
Judgment
- NOT patentable.
Reasoning
- Discovery of fetal DNA = natural phenomenon
- Steps used conventional techniques
Principle
- Even groundbreaking discoveries can be ineligible.
Relevance to AI-Bioinformatics
- AI detecting disease from biological signals:
- If based on natural correlation → fails eligibility
👉 Highly relevant for public health diagnostics AI.
6. Cleveland Clinic v. True Health Diagnostics (2017)
Facts
- Diagnostic method using biomarker (MPO) for heart disease.
Judgment
- NOT patentable.
Reasoning
- Correlation between MPO and disease = natural law
- Testing methods were routine
Principle
- Diagnostic claims remain highly vulnerable post-Mayo.
Relevance
- AI predicting cardiovascular risk from biomarkers:
- Likely ineligible unless novel technical implementation exists
7. Electric Power Group v. Alstom (2016)
Facts
- System for monitoring power grid using data analysis.
Judgment
- NOT patentable.
Reasoning
- Collecting, analyzing, displaying data = abstract idea
Principle
- Data analytics alone is not patentable.
Relevance to AI-Bioinformatics
- AI public health dashboards:
- Disease surveillance systems
- Epidemiological analytics
👉 Must show technical improvement, not just analysis
8. Thaler v. USPTO / DABUS Cases (Global AI Inventorship Cases)
Facts
- AI system (DABUS) listed as inventor.
Judgment
- Rejected (US, UK, EU, etc.)
Principle
- Only natural persons can be inventors
Relevance
- In AI-bioinformatics:
- Human involvement is mandatory
- Fully autonomous AI inventions → not patentable
III. Application to Hybrid AI–Bioinformatics in Public Health
A. Typical Patent-Ineligible Scenarios
- AI predicting disease from biomarkers
- Genomic correlation models
- Epidemiological data analysis systems
👉 Why?
- Natural law + abstract idea combination
- No “inventive concept”
B. Potentially Patent-Eligible Scenarios
A hybrid AI-bioinformatics invention MAY be patentable if it includes:
1. Technical Improvement
- New neural network architecture improving genomic processing
2. Transformation of Data
- Converting raw biological signals into engineered datasets
3. Integration with Hardware
- AI embedded in diagnostic devices (e.g., biosensors)
4. Novel Treatment Methods
- AI-guided therapy (more likely patentable than diagnostics)
C. Key Legal Trend
- Diagnostics → hardest to patent
- Treatment methods → more likely patentable
- AI algorithms → must show technical effect
👉 Courts demand “something more” than:
- natural correlation
- generic AI
- routine lab steps
IV. Critical Evaluation
Strengths of Current Law
- Prevents monopolization of:
- natural biological knowledge
- fundamental algorithms
Weaknesses
- Discourages innovation in:
- public health diagnostics
- AI-driven genomics
- Creates uncertainty:
- hard to distinguish “application” vs “discovery”
V. Conclusion
Patent eligibility of hybrid AI-bioinformatics technologies is governed by a strict judicial framework shaped by cases like:
- Mayo (laws of nature)
- Myriad (natural phenomena)
- Alice (abstract ideas)
Together, they impose a high threshold:
👉 An invention must:
- go beyond biological discovery, AND
- go beyond AI abstraction, AND
- demonstrate a technical inventive concept
RELATED Blog
comments