Chapter Four · failure evidence
What RNA Sequencing & Transcriptomics got wrong, from 49 dissertations
The records document recurring technical and analytical challenges encountered across bulk, single-cell, and spatial transcriptomics studies. Researchers faced significant obstacles including low starting RNA yields, batch effects, spatial deconvolution inaccuracies, statistical power loss, and poor concordance with orthogonal validation assays. These records come from PhD theses at 12 institutions, 2021 to 2026. Each links to its thesis. They were extracted by language models reading the full text, so treat each as a lead to read, not a verdict.
Low starting RNA biomass and poor capture efficiency cause library preparation failure and loss of target transcripts
Experiments involving low-input cells, plasma cell-free RNA, or intracellular parasites frequently failed because insufficient RNA yields or low target read counts prevented successful sequencing and transcript reconstruction. Single-cell and spatial sequencing protocols also suffered from low capture efficiency and failed sample quality control, leading to excluded samples or missing transcripts despite positive targeted detection.
Tried and failed
lentiviral barcode capture via single-cell RNA sequencing applied to differentiated human pluripotent stem cells. Outcome: no signal. Reason: EF-1alpha promoter driven barcoded transcripts were not captured in scRNA-seq despite positive qPCR detection
Lost to a baseline
A considerable number of tissue samples failed RNA quality control criteria prior to small RNA library preparation and were excluded from final microRNA sequencing.
Lost to a baseline
Single-cell RNA sequencing captured ~10% of mRNA, whereas targeted ExSeq yielded ~62% relative to smFISH/ExFISH.
Spatially precise in situ transcriptomics in intact biological systems · MIT
Considered and rejected
Considered and rejected: Single-nuclei RNA sequencing (snRNAseq) was rejected in favour of scRNAseq due to snRNAseq's known under-representation of immune cell populations and exclusion of cytoplasmic RNA.
Elucidating the role of inflammatory mediators in frozen shoulder · Oxford
Lost to a baseline
Slide-seq (version 1) had approximately 2.7% the capture efficiency of Drop-seq single-cell RNA sequencing.
Technologies for assaying the spatial position of biomolecules in situ · Harvard
Tried and failed
nuclear run-on nascent RNA sequencing using HPDP-biotin applied to low-input mammalian cell samples. Outcome: no signal. Reason: insufficient starting cell number resulted in undetectable RNA yield and poor library quality
NAB2-STAT6 DRIVES AN EGR1-DEPENDENT NEUROENDOCRINE PROGRAM IN SOLITARY FIBROUS TUMORS · Penn
Tried and failed
small RNA sequencing for biomarker discovery applied to circulating cell-free RNA in plasma. Outcome: no signal. Reason: insufficient differentially expressed candidates detected in systemic circulation compared to local biofluids
MicroRNAs in extracellular vesicles as biomarkers for pancreaticobiliary cancers · Imperial
Lost to a baseline
7 patient samples (out of 16 sorted pairs) excluded from paired transcriptomic analysis due to low macrophage cell yield (<5000 cells), inadequate RNA content normalisation factor, or outlier status on principal component analysis.
Tried and failed
hybrid-capture RNA sequencing de novo assembly applied to low-biomass intracellular parasite transcriptome. Outcome: data insufficient. Reason: hybrid capture yielded insufficient read coverage to reconstruct target transcripts de novo
Tried and failed
differential expression analysis on low-count conditions applied to bacterial transcriptomes across environmental treatments. Outcome: data insufficient. Reason: statistical robustness was insufficient due to low target gene read counts in experimental condition
Spatial transcriptomics workflows suffer from deconvolution inaccuracies, spot assignment errors, and imaging artifacts
Spot-level deconvolution algorithms and hard class assignments yielded inaccurate cell abundance estimates, while spatial differential expression baseline methods failed to separate cell proportions from within-phenotype expression variation. Spatial analyses also suffered from image blurriness artifacts misleading domain clustering, noisy signals from spatial graph filtering, and transcript coverage depths inferior to targeted dissection methods.
Tried and failed
reference-based transcriptomic deconvolution algorithms applied to spatial transcriptomics cell abundance estimation. Outcome: did not generalise. Reason: algorithms yielded inaccurate and inconsistent abundance estimates when benchmarked against ground-truth spatial proteomics single-cell data
Tried and failed
hard assignment of dominant class labels applied to heterogeneous multi-cell spatial transcriptomics spots. Reason: unaccounted cellular spillover and contamination substantially inflated type I error rates in downstream differential testing
ULTRA SCALABLE METHODS FOR DIFFERENTIAL TESTING OF SPATIAL-OMIC DATA · Penn
Tried and failed
spatial differential expression baseline methods applied to spatially resolved transcriptomics data. Reason: failed to disentangle cell phenotype proportions from within-phenotype spatial gene expression variation
Generative modeling of the tumor microenvironment : deconvolution, completion, and integration · UT Austin
Tried and failed
PCA on graph high-pass filtered spatial signals applied to spatial transcriptomics data. Outcome: no signal. Reason: tissue-wide spatial averaging yielded noisy, non-localized patterns across the whole tissue
Fundamental representations of regions and interactions in spatial transcriptomics · MIT
Tried and failed
geometric sketching and random subsampling applied to spatially resolved transcriptomics data. Outcome: data insufficient. Reason: Oversamples dominant cell populations and misses rare spatial cell configurations and interactions.
ULTRA SCALABLE METHODS FOR DIFFERENTIAL TESTING OF SPATIAL-OMIC DATA · Penn
Tried and failed
dissociated single-cell RNA sequencing applied to spatially localized regional gene expression differences. Outcome: no signal. Reason: loss of spatial context during tissue dissociation obscured sharp anatomical boundary marker expression
SPATIOTEMPORAL PROFILING OF GENE EXPRESSION IN HEALTH AND DISEASE · Cornell
Tried and failed
multimodal clustering with image and transcriptomics data applied to tissue spatial domain identification. Reason: optical blurriness artifacts in images misled clustering away from true biological domains
Machine learning methods for the analysis of multi-modal spatial omics data · Penn
Lost to a baseline
Spatial PCA was beaten in cross-validated MSPE by RapPCA (20.56 vs 46.09, and 49.90 for built-in) and in TMSE (271.35 vs 292.78, and 296.59 for built-in) on HER2+ breast tumor spatial transcriptomics data
Statistical Machine Learning for Spatial- and Network-Linked Data · ResearchWorks
Considered and rejected
Considered and rejected: Rejected using spatial transcriptomics in place of LCM-RNA-seq due to lower depth of transcript coverage.
Volatile hypoxia signatures in oesophageal adenocarcinoma · Imperial
Technical batch effects and biological confounders distort single-cell clustering and integration
Single-cell clustering and dimensionality reduction were frequently misled when global gene sets captured cell cycle or metabolic timepoint differences rather than cell differentiation states. In addition, uncorrected batch effects, misannotated genes driving false clusters, and uniform filtering thresholds that discarded valid low-output cells prevented reliable cell type annotation and cross-sample integration.
Tried and failed
Single-cell RNA sequencing unsupervised clustering applied to Single-cell gene expression profiles. Reason: Misannotated genes artifactually drove false cluster segregation
Investigation of maturation and survival of human long-lived plasma cells using integrated single-cell analysis · Georgia Tech
Tried and failed
Uniform quality filtering thresholds across heterogeneous samples applied to single-cell RNA sequencing data. Reason: Discarded valid cell subpopulations with naturally lower transcriptional output and baseline counts
Considered and rejected
Considered and rejected: Rejected dimensionality reduction (UMAP) on whole/new transcriptomes using all detected genes because it segregated cells by sampling timepoint rather than differentiation state due to cell cycle/metabolism confounding; restricted to T cell differentiation TF module.
Molecular regulation of CD8+ T cell stemness and function · ResearchWorks
Tried and failed
reference mapping algorithms for cell type annotation applied to single-cell transcriptomic data integration. Outcome: did not generalise. Reason: insufficient resolution without artifactual batch-integration forcing
Tried and failed
single-cell RNA sequencing applied to memory T cell subpopulations. Outcome: no signal. Reason: technical batch effects and severe underrepresentation of one target subpopulation confounded clustering and differential expression
Investigating the heterogeneity of the human immune memory compartment · Imperial
Considered and rejected
Considered and rejected: Single-cell transcriptomics without feature barcoding, rejected due to technical noise and severe batch effects across multiple knockout populations
Functional impact of inactivating mutations in epigenetic regulators in cancer · Imperial
Considered and rejected
Considered and rejected: Single-cell transcriptomic analyses performed per individual sample independently were rejected because inter-sample biological variation prevented drawing generalisable cross-sample conclusions.
Molecular characterisation of ASXL1-mutant clonal haematopoiesis · Oxford
Single-cell differential expression analyses suffer from statistical underpowering and inflated false discoveries
Low biological replicate numbers, high within-condition biological variance, and low recovered cell counts severely underpowered pseudobulk and single-cell differential expression testing. Furthermore, single-cell gene dropout obscured tissue-level changes, omitting donor random effects led to massive false positive rate inflation, and cross-cohort tests produced overwhelming gene lists with poor generalizability.
Tried and failed
differential expression analysis on single-cell RNA sequencing applied to whole tissue disease profiling. Outcome: no signal. Reason: gene dropout effects in single-cell sequencing obscured differences across the whole tissue level
Tried and failed
pseudobulk differential expression analysis applied to single-cell RNA sequencing data. Outcome: data insufficient. Reason: low recovered cell counts per cell type across participant groups
Single-Cell Biology of Respiratory Viral Infections in the Nasal Mucosa · Harvard
Tried and failed
differential expression with small sample sizes applied to single-cell RNA sequencing data. Outcome: data insufficient. Reason: low biological replicate counts cause severe statistical underpowering and very high false discovery rates
A novel approach to power analysis for differential expression in scRNA-seq data · Imperial
Tried and failed
differential expression testing without donor random effects applied to single-cell RNA sequencing data. Reason: omitting donor/batch random effects led to massive inflation of false positive rates (type I error)
Cell states and neuronal vulnerabilities in neurodegenerative diseases · Harvard
Tried and failed
pseudobulk differential expression analysis applied to single-nucleus RNA sequencing in animal disease models. Outcome: no signal. Reason: high within-condition biological variance coupled with limited biological replicates underpowered statistical detection
Single-cell RNA-sequencing in epilepsy; discovery of cell-types, pathways, and drug targets · Imperial
Tried and failed
differential expression analysis across independent cohorts applied to cross-dataset transcriptomic disease biomarker discovery. Outcome: did not generalise. Reason: yielded an overwhelming mass of significant genes with poor explainability and lack of generalizability
Improvements in the Modeling of High Dimension/Low Sample Size Imbalanced Clinical Datasets · Georgia Tech
Transcriptomic expression changes fail to correlate with protein abundance or targeted validation assays
Differentially expressed transcripts identified through sequencing frequently failed to replicate when tested against orthogonal assays such as digital droplet PCR, reporter gene expressions, or single-molecule in situ hybridization. Additionally, pronounced transcriptional upregulations failed to produce corresponding changes in protein abundance when evaluated by proteomic profiling or intracellular flow cytometry.
Tried and failed
validating targeted PCR expression via global RNA-sequencing applied to differential gene expression analysis. Outcome: did not generalise. Reason: significant expression differences detected by ddPCR were not detected as differentially expressed in global transcriptomic sequencing
Examining the role of metabolism in CD4+ T cell-driven inflammation · UT Austin
Tried and failed
cross-platform validation of transcriptomics with proteomics applied to differential gene expression analysis. Outcome: did not generalise. Reason: significant mRNA differential expression failed to correlate with corresponding protein abundance changes in proteomic assays
Lost to a baseline
CeNGEN single-cell RNA sequencing profiling did not correlate well with in vivo transcriptional/translational reporter gene expression levels across glr-1 neurons
Behavior-Coupled Neural Circuit Analysis of Chemosensory Responses in C. elegans · Georgia Tech
Tried and failed
intracellular flow cytometry for surface/secreted protein applied to antigen-specific T cell protein expression validation. Outcome: no signal. Reason: Protein level changes were undetectable despite strong transcriptomic upregulation
Mechanisms of T cell dysfunction during human T cell leukemia virus type 1 infection · Imperial
Tried and failed
spatial transcriptomics for differential gene expression applied to spinal tissue sections. Outcome: no signal. Reason: candidate gene downregulation identified by spatial sequencing was not confirmed by single-molecule in situ hybridization
Exploring genetic interactions in adolescent idiopathic scoliosis · UT Austin
Heterogeneous tissue bulk sequencing dilutes and masks cell-type-specific transcriptional signals
Whole-tissue bulk transcriptomic profiling failed to detect cell-type-specific differential expression, rare subclonal relapse pathway enrichment, or mild behavioral stimulation responses because target signals were diluted within cell population averages. Similarly, whole mRNA sequencing was rejected in favor of targeted capture because non-disease background transcripts dominated sequencing depth over splice junctions of interest.
Tried and failed
bulk RNA sequencing applied to detecting cell-type-specific differential expression. Outcome: no signal. Reason: cell-type-specific transcriptional signals were diluted and obscured within heterogeneous tissue samples
Dissecting cellular heterogeneity in human pathologies using single cell genomics · UT Austin
Tried and failed
bulk transcriptomic profiling applied to detecting subclonal relapse pathway enrichment. Outcome: no signal. Reason: signal from rare subclonal populations was diluted and masked in bulk cell population averages
Dissecting cellular heterogeneity in human pathologies using single cell genomics · UT Austin
Tried and failed
bulk RNA sequencing after mild behavioral stimulation applied to heterogeneous brain tissue. Outcome: no signal. Reason: Cellular heterogeneity and mild stimulus strength diluted transcriptional changes below detection thresholds
Considered and rejected
Considered and rejected: Rejected whole mRNA sequencing in favor of targeted capture panels due to high mRNA sequencing cost, read-depth dominance by non-disease transcripts, and insufficient splice junction depth.
Clinical utility of targeted RNAseq in neuromuscular and immune disorders · Georgia Tech
Left open by the authors
Problems the authors named and did not get to.
Left open
Validate RNA-seq transcriptomic findings using secondary assays such as qPCR or in situ hybridization. Blocker: Requires wet lab facilities, tissue samples, and experimental assays (qPCR or in situ hybridization)
Mechanisms Driving the Skeletal Response to Mechanical Loading and Parathyroid Hormone · Cornell
Left open
Validate candidate genes from the transcriptomics study, such as AHSP and TFR2, using qPCR. Blocker: Requires a wet lab, biological samples, and qPCR equipment.
Muscle wasting in older adults · Imperial
Left open
Validate inferCNV-derived subclonal copy number variation profiles from single-nucleus RNA sequencing using matched patient whole-genome sequencing data. Blocker: Requires matched patient whole-genome sequencing data from the private tumor cohort or wet-lab sequencing of patient samples
Multiomic insights into gastroenteropancreatic neuroendocrine tumor biology · Harvard
Left open
Integrate single-cell RNA-seq or spatial transcriptomics to validate ARF-induced gene expression changes in specific cell types. Blocker: Requires wet-lab experiments to generate or validate single-cell/spatial transcriptomics data on engineered plant lines
Decoding the Transcriptional Specificity of Auxin Signaling: A Synthetic Biology Approach · Virginia Tech
Left open
Use optimized triple-reporter flow cytometry to profile the transcriptome and chromatin of switched IFNg+ Th1-like cells. Blocker: Requires wet lab facilities, specific reporter mice/cells, flow cytometry cell sorting, and sequencing infrastructure.
INTERLEUKIN 2 INDUCIBLE T CELL KINASE (ITK) FINE TUNES T CELL DIFFERENTIATION · Cornell
Left open
Disentangle lineage relationships between early effector and chronic exhausted CD8+ T cell subsets using scRNA-seq trajectory analysis, spatial transcriptomics, or CRISPR barcoding. Blocker: Requires wet lab experimentation involving single-cell spatial transcriptomics, single-cell RNA-seq, or CRISPR barcoding in infection models.
Cell Intrinsic Factors Influence CD8+ T Cell Fate Decisions Following Acute and Chronic Infection · Cornell
Left open
Assess cellular resource competition effects on nuclear import/export, folding, and post-translational machinery using transcriptomic and proteomic profiling. Blocker: Requires wet lab experimental generation of transcriptomic and proteomic data from mammalian cell cultures.
Resource-aware mammalian cell engineering · Imperial
Left open
Evaluate and characterize non-dichotomized B cell helper T cell populations beyond classical Tfh and Tph using single-cell RNA sequencing data. Blocker: Requires murine wet-lab experiments, FACS sorting, and sequencing or access to unpublished raw single-cell sequencing datasets.
Peripheral T helper Cells in a Murine Chronic Inflammation Model · Harvard
Left open
Perform low-input RNA-sequencing and transcriptomic analysis on sorted GFP+ versus mCherry+ disseminated tumor cells from Jedi mice. Blocker: Requires wet-lab sorting of mouse tissue and low-input RNA sequencing experimental protocols.
Characterization of Immune Evasion Mechanisms of Breast Cancer Disseminated Cells in the Lung · Harvard
Left open
Perform spatial and single-cell transcriptomics on PLD-high and PLD-low xenograft cells sorted by IMPACT to assess metastasis capability. Blocker: Requires wet lab facilities, animal xenograft models, cell sorting via IMPACT, and transcriptomic sequencing apparatus
PHOTOAFFINITY LABELING STRATEGIES TO STUDY PHOSPHOLIPASE D SIGNALING · Cornell
Checking a claim in this area?
We can run the same search on any method or claim. If nothing turns up, we will say so, and that proves nothing on its own.