Scholarlink

<p>Personalized gene editing demands robust mechanisms for privacy, ethical governance, and verifiable data integrity. This paper proposes ViBioChain, a modular blockchain-anchored architecture integrating five components: (1)&#xa0;differential chain-of-custody audit combining quantum fingerprinting with post-quantum signatures for immutable genomic audit trails; (2)&#xa0;proof-of-bioethical-compliance employing zero-knowledge proofs and AI-based ontology evaluation for automated bioethical gating; (3)&#xa0;federated genomic trust mesh (FGTM) enabling privacy-preserving collaborative model training with Renyi differential privacy accounting and trust-weighted federated aggregation; (4)&#xa0;ethical smart orchestration network for modular smart-contract-based workflow governance; and (5)&#xa0;genomic impact estimator via ethical explainability graphs (GIE-EEG) for ancestry-aware, ethically constrained phenotypic forecasting. Afterexpert-driven reconciliation, the implementation was rerun using 800 simulated individuals per dataset, 120 binary loci, five institutional clients, five independent seeds (42–46), and a true trust-weighted federated logistic aggregation path for FGTM rather than the earlier centralized accuracy proxy. Across three genomic cohorts and three domain-comparable baselines, ViBioChain achieved 92.16% ethical violation interception, 100.00% audit trail accuracy, 99.47% workflow traceability, 0.9183 ethical score alignment, and the highest global model accuracy among the tested methods (74.36%). The formal Renyi differential privacy accountant remained within budget (<InlineEquation ID="IEq1"><EquationSource Format="TEX">\(\varepsilon =1.8578\)</EquationSource></InlineEquation>, <InlineEquation ID="IEq2"><EquationSource Format="TEX">\(\delta =10^{-5}\)</EquationSource></InlineEquation>); however, the conservative clean-versus-noisy update leakage proxy did not support the earlier lowest-empirical-leakage assertion. That claim has therefore been removed. Additional IID and non-IID experiments show that severe Dirichlet client heterogeneity (<InlineEquation ID="IEq3"><EquationSource Format="TEX">\(\alpha =0.1\)</EquationSource></InlineEquation>) reduced final accuracy by 1.70–4.10 percentage points relative to IID partitions. The revised results provide a more conservative and reproducible blueprint for secure, ethically governed, and explainable genomic medicine in multi-institutional settings.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ViBioChain: a blockchain-enabled architecture for privacy-preserving, ethically governed, and explainable personalized gene editing

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月18日

查看全文
  • C. Prabakaran,
  • R. Kannadasan

摘要

Personalized gene editing demands robust mechanisms for privacy, ethical governance, and verifiable data integrity. This paper proposes ViBioChain, a modular blockchain-anchored architecture integrating five components: (1) differential chain-of-custody audit combining quantum fingerprinting with post-quantum signatures for immutable genomic audit trails; (2) proof-of-bioethical-compliance employing zero-knowledge proofs and AI-based ontology evaluation for automated bioethical gating; (3) federated genomic trust mesh (FGTM) enabling privacy-preserving collaborative model training with Renyi differential privacy accounting and trust-weighted federated aggregation; (4) ethical smart orchestration network for modular smart-contract-based workflow governance; and (5) genomic impact estimator via ethical explainability graphs (GIE-EEG) for ancestry-aware, ethically constrained phenotypic forecasting. Afterexpert-driven reconciliation, the implementation was rerun using 800 simulated individuals per dataset, 120 binary loci, five institutional clients, five independent seeds (42–46), and a true trust-weighted federated logistic aggregation path for FGTM rather than the earlier centralized accuracy proxy. Across three genomic cohorts and three domain-comparable baselines, ViBioChain achieved 92.16% ethical violation interception, 100.00% audit trail accuracy, 99.47% workflow traceability, 0.9183 ethical score alignment, and the highest global model accuracy among the tested methods (74.36%). The formal Renyi differential privacy accountant remained within budget (\(\varepsilon =1.8578\), \(\delta =10^{-5}\)); however, the conservative clean-versus-noisy update leakage proxy did not support the earlier lowest-empirical-leakage assertion. That claim has therefore been removed. Additional IID and non-IID experiments show that severe Dirichlet client heterogeneity (\(\alpha =0.1\)) reduced final accuracy by 1.70–4.10 percentage points relative to IID partitions. The revised results provide a more conservative and reproducible blueprint for secure, ethically governed, and explainable genomic medicine in multi-institutional settings.