Overcoming data privacy and data gravity challenges in bioinformatics research
Bibliographic record
Abstract
Next-generation sequencing technologies have generated a massive amount of DNA, RNA, and protein sequences since their inception. However, data privacy policies often restrict sharing such data for the risk of re-identifying individuals from whom the sequences were generated. Even when all the data from a sequencing experiment is available, it is often insufficient for statistical power or training machine learning models. Despite the lack of data, sometimes the data sets are ironically too large to realistically share with researchers. In this thesis, I explore methods to overcome challenges of data privacy and data gravity in bioinformatics research. In collaboration with QIMR Berghofer and the Riken Center for Integrative Medical Sciences, we used federated methods to analyze genomic data from the BioBank Japan in situ to classify variants of uncertain significance while preserving privacy. With the Department of Laboratory Medicine and Pathology at the University of Washington, we developed a statistical model that demonstrates how using responsibly shared clinical evidence alone can classify variants of uncertain significance which occur at the rate of 1 in 100,000 people within just a few years. With researchers from McGill University, we reviewed the state of the art in federated computing technologies and how well they satisfy the privacy restrictions from the General Data Protection Regulation. With researchers from NASA, Amazon, and Intel, we developed a federated learning framework to run between terrestrial and space-borne compute infrastructure, laying the groundwork for subsequent experiments which preclude the need to transfer large datasets across astronomical distances. Finally, at NASA, we used a causal inference machine learning ensemble to infer robust correlation between mouse liver gene expression and a corresponding lipid density phenotype in space-flown mice.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.150 | 0.247 |
| Meta-epidemiology (narrow) | 0.001 | 0.002 |
| Meta-epidemiology (broad) | 0.003 | 0.003 |
| Bibliometrics | 0.004 | 0.009 |
| Science and technology studies | 0.005 | 0.020 |
| Scholarly communication | 0.015 | 0.039 |
| Open science | 0.006 | 0.020 |
| Research integrity | 0.007 | 0.016 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".