Rxivist logo

Privacy Preserving RNA-Model Validation Across Laboratories

By Talal Ahmed, Mark A Carty, Stephane Wenric, Jonathan R Dry, Ameen Abdulla Salahudeen, Aly A. Khan, Eric Lefkofsky, Martin C. Stumpe, Raphael Pelossof

Posted 04 Apr 2021
bioRxiv DOI: 10.1101/2021.04.01.437893

Reproducibility of results obtained using RNA data across labs remains a major hurdle in cancer research. Often, molecular predictors trained on one dataset cannot be applied to another due to differences in RNA library preparation and quantification. While current RNA correction algorithms may overcome these differences, they require access to patient-level data which carries inherent risk of loss of privacy. Here, we describe SpinAdapt, a novel unsupervised domain adaptation algorithm that enables the transfer of molecular models across laboratories without access to patient-level sequencing data thereby minimizing privacy risk. SpinAdapt computes data corrections via aggregate statistics of each dataset, rather than requiring full sample-level data access, thereby maintaining patient data privacy. Furthermore, decoupling the model from its training data allows the correction of new streaming prospective data, enabling model evaluation on validation cohorts. SpinAdapt outperforms current correction methods that require patient-level data access. We expect this novel correction paradigm to enhance research reproducibility, quality, and patient privacy.

Download data

  • Downloaded 150 times
  • Download rankings, all-time:
    • Site-wide: 125,499
    • In bioinformatics: 10,037
  • Year to date:
    • Site-wide: 28,570
  • Since beginning of last month:
    • Site-wide: 32,095

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide