Skip to main content
2024 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

High School Poster

HS Poster

Poster 25 - Applying the DeepSeqDock Harmonization Framework on Multi-Institutional Universal Human Reference RNA

Saturday, October 26, 2024
10:00 AM - 12:00 PM EST
Location: Exhibit Hall E, F & G

Introduction: In the ever-increasing availability of public RNA sequencing data, there have been many efforts to reduce batch effects through data harmonization to remove confounding variables in downstream interpretations while allowing the pooling of data for larger sample sizes and stronger analysis. One such pipeline created is DeepSeqDock [1], a framework developed on sequencing quality control (SEQC) data [2]. In this FDA-led global project, various worldwide labs generated RNA-sequencing data to test for reproducibility and transferability, containing Universal Human Reference RNA, Universal Human Brain RNA, and datasets created with ratios of both. In this study, we adjusted and adapted the framework to handle non-SEQC data, namely multi-institutional Universal Human Reference RNA samples. By implementing data harmonization, we simultaneously increase the sample size by pooling data and removing technical noise, allowing for a more powerful analysis to capture the value of a global data network. This study shows a generalization of the DeepSeqDock framework to adapt to data from online sources, like the National Center for Biotechnology Information (NCBI) database [3], a demonstration that simplifies future generalized data harmonization through the adapted framework.

Materials and

Methods: We examined data from the NCBI for gene-expression analysis across studies and chose to focus on the Universal Human Reference RNA [4]. Data from the SEQC was parsed, and samples derived from the Universal Human Reference RNA were used to generate a JavaScript Object Notation (JSON) file containing harmonization parameters using the DeepSeqDock. We focused on GSM749465 [5] and PRJNA208369 [6] for our external datasets, two distinct institutions that performed genome analysis. Data was downloaded using the ‘prefetch’ and ‘fastq-dump’ [7] tools, and then compressed with the ‘gzip’ tool. Using a script for uniform ARCHS4 [8] preprocessing, we re-aligned the FASTQ files into a comma-separated values (CSV) format. This CSV was then transformed and harmonized with frozen parameters from the DeepSeqDock JSON file inside the DeepSeqDock docker container environment.

Results, Conclusions, and Discussions: We successfully applied the DeepSeqDock framework to harmonize these two datasets, using the parameters derived from the SEQC samples for only the Universal Human Reference RNA [4]. We were able to reduce batch effects and noise from multi-institutional technical factors. The harmonization process created a transformed dataset, in which we performed a similarity comparison and generated a heatmap image for visualization, principal component analysis graphs, and results from the evaluation metrics.

This study successfully adapted the DeepSeqDock [1] framework for multi-institutional data not limited to the SEQC dataset [2], highlighting the broader potential and applications of data harmonization. Future work includes generalizing towards bacteria and non-human species.

Acknowledgements (Optional): Funding was provided by the Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health (P20GM103443). We thank the USD Biomedical Engineering Department for their guidance, mentorship, feedback, and support on this project.

References (Optional): [1] bicbioeng, “GitHub - bicbioeng/DeepSeqDock: Framework for building optimized autoencoders for RNA sequencing data harmonization.,” GitHub, 2022. https://github.com/bicbioeng/DeepSeqDock (accessed Aug. 15, 2024).

[2] “A comprehensive assessment of RNA-seq accuracy, reproducibility and information content by the Sequencing Quality Control Consortium,” Nature Biotechnology, vol. 32, no. 9, pp. 903–914, Aug. 2014, doi: https://doi.org/10.1038/nbt.2957.

[3] National Center for Biotechnology Information, “Sequence Read Archive,” National Institutes of Health, 2024. https://www.ncbi.nlm.nih.gov/sra (accessed Aug. 15, 2024).

[4] National Center for Biotechnology Information, “GEO Accession viewer - GSM967566,” National Institutes of Health, 2024. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM967566 (accessed Aug. 15, 2024).

‌[5] National Center for Biotechnology Information, “GEO Accession viewer - GSM749465,” National Institutes of Health, 2024. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM749465 (accessed Aug. 15, 2024).

[6] National Center for Biotechnology Information, “A comprehensive assessment of RNA-seq accuracy, re... (ID 208369) - BioProject - NCBI,” National Institutes of Health, 2014. https://www.ncbi.nlm.nih.gov/bioproject/208369 (accessed Aug. 15, 2024).

[7] National Center for Biotechnology Information, “SRA Tools,” National Institutes of Health, 2024. https://github.com/ncbi/sra-tools/wiki (accessed Aug. 15, 2024).

[8] A. Lachmann et al., “Massive mining of publicly available RNA-seq data from human and mouse,” Nature Communications, vol. 9, no. 1, Apr. 2018, doi: https://doi.org/10.1038/s41467-018-03751-6.