T2D Bulk Blood Transcriptome Signal-to-Noise
Bulk whole-blood RNA-seq in T2D faces a fundamental signal-to-noise challenge: the disease-associated transcriptomic signal is weak relative to variance driven by other subject-level characteristics.
Tkachenko et al. 2024 Evidence
Tkachenko et al. 2024 performed bulk RNA-seq on whole blood from 9 T2D and 9 controls (Russian cohort). Key results demonstrating the signal-to-noise challenge:
- Only 146 DEGs (0.2% of 74,163 expressed genes) at adjusted p < 0.05 — 71 up, 75 down.
- PCA biplot using the top 500 most variable genes shows samples mostly mix and do not group by disease status. The authors explicitly state “the majority of variance in the experiment is not explained by the disease status and might be attributed to some other characteristics describing subjects.” ^[extracted] (from Tkachenko et al. 2024)
- Enriched GO terms were broad: RNA splicing, proteasomal catabolism, chromatin binding, T cell activation — processes that could be affected by many non-disease factors.
Implications for Blood Transcriptomics in T2D
- The low signal-to-noise ratio helps explain why individual blood transcriptomic studies in T2D often show poor cross-study concordance.
- Bulk blood transcriptomic findings are sensitive to cohort composition, sample handling, sequencing depth, and statistical power.
- Meta-analysis approaches can recover weaker coordinated signals across studies, as demonstrated by Tkachenko et al. 2025, but the fundamental variance structure constrains what individual blood transcriptomic studies can detect.
Relevance to Project
- The paper’s PBMC multi-omic data (scRNA-seq, scATAC-seq) should be expected to have better signal-to-noise than bulk blood because it resolves cell-type-specific expression and avoids whole-blood heterogeneity.
- However, if the project generates its own bulk blood RNA-seq data for comparison, the weak disease-variance signal should be anticipated and power calculations should account for it.