Federated learning models for genomic research

Introduction

Federated Learning (FL) models for genomic research refer to decentralised machine learning systems that allow multiple research institutions, hospitals, laboratories and biotechnology organisations to collaboratively analyse genomic data without directly transferring their raw datasets to a central location.

Genomic research involves the study of DNA sequences, genetic variations, inherited diseases and biological characteristics. Such research requires access to large quantities of genetic information, which may contain sensitive personal and familial data. Traditional research models often require centralised data collection, creating concerns relating to privacy, confidentiality, cybersecurity and regulatory compliance.

Federated learning provides an alternative approach by allowing participating institutions to train machine learning models locally and share model updates rather than routinely transferring raw genomic records. This approach can support collaborative disease research, genetic diagnosis, drug discovery and personalised medicine.

However, federated learning does not automatically eliminate privacy risks. Model updates may still reveal sensitive information, making appropriate legal safeguards, cybersecurity controls and ethical supervision necessary.

Meaning and objectives of federated learning

Federated learning is a machine learning technique in which several participants collaboratively train a common model while retaining their original datasets within their respective institutions.

For example, three hospitals conducting genetic research may independently train the same machine learning model using their genomic datasets. They can exchange model updates with a coordinating server, which combines the updates and distributes an improved model.

The principal objectives include:

Protecting the confidentiality of genomic information.

Facilitating international genomic research collaboration.

Reducing unnecessary transfers of sensitive genetic data.

Improving the accuracy of predictive research models.

Supporting research involving rare genetic diseases.

Encouraging cooperation between hospitals and biotechnology institutions.

Promoting responsible use of artificial intelligence in healthcare.

Major federated learning models used in genomic research

1. Horizontal federated learning

Horizontal federated learning is used when different institutions possess similar types of genomic information relating to different individuals.

For example, multiple hospitals may have genetic information concerning patients with similar diseases but different demographic characteristics.

Each institution trains the model using its local dataset and shares model updates for collaborative learning.

2. Vertical federated learning

Vertical federated learning is used when different institutions possess different types of information relating to overlapping individuals.

For example, one institution may possess genomic information, while another holds relevant clinical or medical records.

The institutions collaboratively develop analytical models without necessarily exchanging their complete datasets.

3. Federated transfer learning

Federated transfer learning is useful when participating institutions have different datasets and limited overlapping information.

It allows knowledge obtained from one dataset to support learning in another institution while maintaining decentralised data arrangements.

4. Federated deep learning

Federated deep learning uses neural networks to analyse complex genomic information.

It can support research involving genetic variants, disease prediction, molecular classification and biological pattern recognition.

5. Secure federated learning

Secure federated learning incorporates additional privacy and security measures, such as:

Secure aggregation.

Encryption.

Differential privacy.

Access controls.

Authentication mechanisms.

Monitoring of model updates.

These measures reduce the risks associated with unauthorised access and disclosure of sensitive genomic information.

Applications of federated learning in genomic research

1. Rare genetic disease research

Rare genetic diseases often involve small patient populations distributed across different geographical regions.

Federated learning allows research institutions to collaborate without requiring all genomic datasets to be transferred to one central database.

This can improve the ability to identify genetic patterns associated with uncommon disorders.

2. Cancer genomics

Cancer research frequently involves analysing genetic mutations and molecular characteristics of tumours.

Federated learning can help hospitals develop predictive models using information from multiple institutions.

It may support research into cancer classification, treatment response and personalised medicine.

3. Drug discovery

Pharmaceutical and biotechnology companies can use federated learning to analyse distributed genomic and clinical datasets.

This can support research into genetic biomarkers, drug responses and potential therapeutic targets.

4. Population genomics

Genetic characteristics may differ across geographical and population groups.

Federated learning can support collaborative population-genetics research while allowing participating institutions to maintain control over their local datasets.

5. Personalised medicine

Federated learning can contribute to predictive models that examine relationships between genetic variations and treatment responses.

Such research may help develop more individualised approaches to medical treatment.

International legal and regulatory framework

1. General Data Protection Regulation (GDPR), 2016

The GDPR establishes a comprehensive framework for processing personal data within the European Union.

Article 4 defines personal data, while Article 9 provides special protection for genetic data and other sensitive categories of personal information.

Article 25 establishes the principle of data protection by design and by default.

Article 32 addresses security measures required for processing personal data.

Federated learning may support GDPR compliance by reducing the need to transfer raw genomic datasets. However, model updates and resulting models must also be assessed for possible personal-data disclosure.

2. UNESCO Universal Declaration on the Human Genome and Human Rights, 1997

The Declaration recognises the human genome as a fundamental element of human identity and emphasises respect for human dignity.

It also addresses the confidentiality of genetic information and the ethical responsibilities associated with genetic research.

These principles are relevant to federated genomic research involving multiple institutions and international participants.

3. International Covenant on Economic, Social and Cultural Rights, 1966

Article 15 recognises the right to enjoy the benefits of scientific progress and its applications.

Federated genomic research can support this objective by facilitating scientific cooperation and expanding access to research findings.

4. OECD Recommendation on Health Data Governance, 2016

The OECD recommendation promotes responsible health-data governance, including privacy protection, security, transparency and international cooperation.

Its principles are relevant to institutions participating in federated genomic research.

Indian legal framework governing federated genomic research

1. Digital Personal Data Protection Act, 2023

India's Digital Personal Data Protection Act establishes a framework for processing digital personal data.

Federated genomic research involving identifiable individuals must be assessed against the Act's applicable requirements concerning lawful processing, security safeguards and individual rights.

The fact that genomic data remains within a participating institution does not automatically remove the processing from the scope of data protection law.

2. Information Technology Act, 2000

The Information Technology Act provides the legal framework for electronic records and cybersecurity-related matters.

Section 43 and Section 66 address certain forms of unauthorised access and computer-related misconduct.

3. Information Technology (Reasonable Security Practices and Procedures and Sensitive Personal Data or Information) Rules, 2011

These Rules historically provided requirements concerning sensitive personal data, including information relating to health conditions and medical records.

Their current application must be considered alongside subsequent amendments and the Digital Personal Data Protection framework.

4. National Ethical Guidelines for Biomedical and Health Research Involving Human Participants, 2017

Issued by the Indian Council of Medical Research, these guidelines establish ethical standards for biomedical and health research involving human participants.

They address informed consent, confidentiality, privacy, research ethics committees and responsible handling of biological and health information.

5. Digital India and National Health Mission initiatives

India's broader digital health initiatives support the development of digital healthcare infrastructure and responsible information management.

However, participation in federated genomic research must also satisfy the applicable privacy, research ethics and institutional requirements.

Important case laws

There are limited judicial decisions directly concerning federated learning in genomic research. Nevertheless, the following judgments establish important legal principles relating to privacy, genetic information, scientific research and data protection.

1. Justice K.S. Puttaswamy (Retd.) v. Union of India (2017)

The Supreme Court of India recognised the right to privacy as a fundamental right under Article 21 of the Constitution.

The judgment discussed informational privacy and the importance of protecting personal information from unjustified interference.

Legal significance: Genomic information can reveal highly sensitive details about individuals and their biological relatives. The judgment provides a constitutional foundation for privacy safeguards in federated genomic research.

2. Justice K.S. Puttaswamy (Retd.) v. Union of India (Aadhaar) (2018)

The Supreme Court examined the constitutional validity of the Aadhaar framework and considered issues relating to personal information, data security and informational privacy.

The Court emphasised the importance of proportionality and safeguards when the state processes personal information.

Legal significance: The decision is relevant to the governance of large-scale genomic databases and decentralised research systems because it highlights the importance of limiting data collection and protecting sensitive information.

3. Selvi v. State of Karnataka (2010)

The Supreme Court examined the constitutional validity of involuntary narco-analysis, polygraph examinations and brain electrical activation profile tests.

The Court held that involuntary administration of these techniques violated constitutional protections under Articles 20(3) and 21.

Legal significance: The judgment reinforces individual autonomy and constitutional protection against certain forms of compelled scientific investigation. These principles are relevant to the ethical use of biological and genetic information.

4. S. and Marper v. United Kingdom (2008)

The European Court of Human Rights examined the retention of fingerprints and DNA samples of individuals who had not been convicted of criminal offences.

The Court found that the blanket retention arrangements violated Article 8 of the European Convention on Human Rights.

Legal significance: The judgment highlights the sensitivity of genetic information and the need for proportionate retention policies. It is relevant to genomic research institutions that maintain long-term genetic datasets.

5. Google Spain SL v. Agencia Española de Protección de Datos (2014)

The Court of Justice of the European Union considered issues relating to personal information, search engine activities and the right to erasure under European data protection law.

Legal significance: Although not a genomic research case, the decision illustrates the importance of individual privacy rights and responsible information management in digital environments.

Legal and ethical challenges in federated genomic research

Federated learning offers important privacy advantages, but it also creates several legal and ethical challenges.

1. Model privacy risks: Shared model updates may reveal information about individuals in the underlying datasets.

2. Informed consent: Research participants must receive appropriate information about how their genomic data will be used in collaborative research.

3. Cross-border data governance: Participating institutions may operate under different national privacy and research regulations.

4. Data ownership and control: Disputes may arise concerning the control, permitted use and commercial exploitation of genomic datasets and trained models.

5. Algorithmic bias: Models trained on unrepresentative population data may produce inaccurate results for certain demographic groups.

6. Institutional accountability: It may be difficult to identify responsibility when an error or privacy breach involves multiple participating institutions.

7. Intellectual property rights: Ownership of collaboratively developed machine learning models may create disputes between hospitals, universities and biotechnology companies.

Recommendations for an effective regulatory framework

An effective governance framework for federated genomic research should include:

Clear informed-consent procedures for genomic research participants.

Secure aggregation and appropriate encryption mechanisms.

Independent research ethics committee approval.

Regular cybersecurity assessments.

Data minimisation and purpose limitation.

Appropriate controls over model updates and trained models.

Transparent agreements between participating institutions.

Procedures for reporting and investigating data breaches.

Fair representation of different population groups.

Clear rules governing intellectual property and commercialisation.

Periodic audits of federated learning systems.

Conclusion

Federated learning models represent an important development in genomic research because they enable collaborative scientific analysis while reducing the need for centralised transfers of sensitive genetic information.

Different models, including horizontal federated learning, vertical federated learning and federated deep learning, provide opportunities for research into rare genetic diseases, cancer genomics, drug discovery and personalised medicine.

International frameworks such as the GDPR and UNESCO Universal Declaration on the Human Genome and Human Rights provide important principles for protecting genetic information. In India, the Digital Personal Data Protection Act, 2023, and ICMR National Ethical Guidelines, 2017, contribute to the regulatory and ethical framework.

Judgments such as Justice K.S. Puttaswamy v. Union of India and S. and Marper v. United Kingdom demonstrate the importance of privacy, proportionality and responsible management of genetic information.

Ultimately, federated genomic research requires a balance between scientific collaboration, individual privacy, cybersecurity and institutional accountability. A strong legal framework can encourage innovation while ensuring that sensitive genomic information is used responsibly and ethically.