JINet: easy and secure private data analysis for everyone

TL;DR

JINet tackles the privacy, interoperability, and accessibility barriers in healthcare data analysis by enabling browser-based, privacy-preserving analyses that run entirely on the client using WebAssembly-powered runtimes (Python via Pyodide and R via WebR). The server acts as an index and distribution hub, transferring application scripts to the client while never accessing raw user data or execution parameters, and enabling encrypted, passphrase-protected sharing of results. The system promotes a self-sustaining ecosystem with roles for Users, Application Developers, and Data Providers, and supports data and tool sharing without compromising confidentiality. This design can significantly enhance reproducibility, security, and broad accessibility for genomic and clinical data analyses, while reducing data leakage risks and vendor lock-in.

Abstract

JINet is a web browser-based platform intended to democratise access to advanced clinical and genomic data analysis software. It hosts numerous data analysis applications that are run in the safety of each User's web browser, without the data ever leaving their machine. JINet promotes collaboration, standardisation and reproducibility by sharing scripts rather than data and creating a self-sustaining community around it in which Users and data analysis tools developers interact thanks to JINets interoperability primitives.

Figures (6)

• Figure 1: A: a figurative representation of the current major burdens encountered in data sharing within medical research. B: a strategic roadmap designed to address the challenges in data sharing and analysis for medical research, creating an environment conducive to robust and secure data-sharing practices. C: JINet ecosystem, facilitating collaboration among users, application developers, and data providers. JINet offers a centralised application distribution platform that enhances data privacy and security, supports versatile data analysis, and promotes community engagement and large-scale collaborative research. The ecosystem includes features such as data analysis tools democratisation, efficient and secure application architecture, and a repository for data formats and applications. The workflow shows the process from application choice to results visualisation and sharing, emphasising ease of use and security in data handling.
• Figure 2: A: A graphical description of the workflow of 3 applications currently available on JINet, adapted from netANOVA duroux2023netanova, netMUG li2023netmug, and GMIC duroux2023graph . B A screenshot from JINet displaying an index of applications. C Running an application is as simple as graphically selecting the parameters to run with them pressing the "Run" button.
• Figure 3: Matrix multiplication in the WASM R interpreter (left) vs. native matrix multiplication with OpenBLAS (right). Each iteration of the benchmark generates 2 random square matrices which are then multiplied with the %*% operator. This benchmark is run 100 times for each matrix size.
• Figure 4: Simulated coin flips in the WASM R interpreter (left) vs. native R interpreter (right). Each iteration of the benchmark generates a large vector containing the string "H" or "T" selected randomly. The number of "H" strings are then counted. This benchmark is run 100 times for each vector size.
• Figure 5: Computation of an inverse matrix from a randomly generated square matrix in the WASM Python interpreter (left) vs. native Python interpreter with OpenBLAS (right). Each iteration of the benchmark generates a random square matrix which is then inverted. This benchmark is run 100 times for each matrix size.