What Epigenetic & Chromatin Profiling got wrong, from 40 theses

Tried and failed

motif-altering regulatory variant mapping to downstream expression applied to transcription factor binding regulatory loci. Outcome: no signal. Reason: chromatin accessibility and histone alterations frequently did not translate into detectable gene expression QTLs

The effects of trait-associated variation on transcription factor binding and other regulatory phenotypes · Harvard · PhD

Tried and failed

correlating differential DNA methylation with differential expression applied to host response to bacterial infection. Outcome: no signal. Reason: differential methylation changes showed no direct overlap with differentially expressed genes

Functional Epigenomics In Insects Using Next-Generation Sequencing Methods · Georgia Tech · PhD

Tried and failed

correlating differential DNA methylation with differential gene expression applied to identifying functionally regulated candidate genes. Outcome: no signal. Reason: very small direct overlap between differentially methylated region genes and differentially expressed genes

Hippocampal Epigenomic and Transcriptional Plasticity of Cocaine Seeking and Extinction · Cornell · PhD

Tried and failed

correlating parent-of-origin differential methylation with gene expression applied to reciprocal cross epigenetic gene regulation. Outcome: no signal. Reason: differential methylation did not correlate with parent-of-origin differentially expressed genes

Functional Epigenomics In Insects Using Next-Generation Sequencing Methods · Georgia Tech · PhD

Tried and failed

chromatin accessibility gene score modeling applied to predicting single-cell gene expression. Outcome: worse than baseline. Reason: distance-weighted accessibility aggregation around gene bodies produced negative out-of-sample R2 scores

Integrating genomic and multiomic data for computational analysis of gene regulation in circulating immune cells · Georgia Tech · PhD

Tried and failed

Gene activity scoring from chromatin accessibility applied to predicting single-cell gene expression. Outcome: no signal. Reason: Accessibility scores correlated poorly with measured RNA levels across multiple scoring algorithms

Integration of single-cell multimodal data to define the chromatin landscape of rheumatoid arthritis · Harvard · PhD

Tried and failed

exponential distance-decay gene activity modeling applied to predicting single-cell gene expression from chromatin accessibility. Outcome: no signal. Reason: modeled accessibility scores correlated poorly with measured mRNA expression levels

Single-Cell Genomics of Human Cancer · Harvard · PhD

Tried and failed

open chromatin peak based enhancer selection applied to cell type specific viral promoter design. Outcome: no signal. Reason: high chromatin accessibility did not correlate with transcriptional activation strength

Engineering enhancers in adeno-associated viral vectors to target specific expression in motor neurons · Harvard · Master's

Considered and rejected

Considered and rejected: Rejected simple correlation between TF expression and motif accessibility in open chromatin; required per-locus linear regression clustering because TF expression did not reliably align with motif accessibility alone.

A Multimodal Molecular View of Human Cartilage Development · Harvard · PhD

Tried and failed

nearest-neighbor feature assignment on 1D genomic sequence applied to predicting gene expression from enhancer acetylation. Outcome: no signal. Reason: linear proximity fails to capture functional distal 3D chromatin interactions regulating target promoters

Multiscale Modeling Of Cell Fate Switching To Predict Patient-Specific Response To Combination Cancer Therapy · Penn · PhD

Tried and failed

using chromatin accessibility to measure promoter inactivation applied to enhancer knockout transcriptional silencing. Outcome: no signal. Reason: promoter accessibility remained unchanged despite near-complete loss of transcription and active histone marks

CRISPR Screening Uncovers Long-Range Enhancers in Pancreatic Differentiation and Links a Diabetes Risk Variant · Cornell · PhD

Tried and failed

epigenomic profiling of regulatory elements applied to rare disease variant enrichment analysis. Outcome: no signal. Reason: chromatin accessibility and histone marks alone insufficiently capture functional disruption by rare non-coding variants

DECIPHERING MULTI-LAYER FUNCTIONAL EFFECTS OF GENOMIC VARIANTS IN HUMAN DISEASES · Cornell · PhD

Tried and failed

cross-cohort differential chromatin peak overlap analysis applied to on-treatment patient epigenomic profiles. Outcome: did not generalise. Reason: Differential histone modification peaks failed to replicate significantly across independent clinical cohorts post-treatment

Computational dissection and prediction of cancer immunotherapy response · MIT · PhD

Tried and failed

gene ontology enrichment on selected regression features applied to blood DNA methylation epigenetic biomarkers. Outcome: no signal. Reason: enrichment terms lost statistical significance after multiple testing correction

DECIPHERING EPIGENETIC SIGNATURES OF AGING AND NEURODEGENERATION IN ALZHEIMER’S DISEASE WITH MACHINE LEARNING · Penn · PhD

Tried and failed

epigenome-wide association study on longitudinal methylation changes applied to longitudinal whole-blood DNA methylation data. Outcome: no signal. Reason: No individual CpG sites reached genome-wide significance after multiple testing correction.

Posttraumatic Stress Disorder in Civilian Women: A Longitudinal Investigation on Cognition, Epigenetics, and Inflammation · Harvard · PhD

Tried and failed

regularized linear regression on high-dimensional epigenetic data applied to predicting brain structural MRI volume. Outcome: did not generalise. Reason: blood epigenetic signatures did not capture brain regional neurodegeneration features across cohorts

DECIPHERING EPIGENETIC SIGNATURES OF AGING AND NEURODEGENERATION IN ALZHEIMER’S DISEASE WITH MACHINE LEARNING · Penn · PhD

Tried and failed

single epigenetic marker for disease classification applied to cancer screening validation cohort. Outcome: did not generalise. Reason: marker achieved chance-level performance (AUC = 0.50) in an independent validation cohort

Genetic, epigenetic and environmental determinants of cervical cancer risk and implications for early detection and screening · Imperial · PhD

Lost to a baseline

The published commercial Infinium Mouse Methylation BeadChip multi-tissue clock (Zhou et al. 2022) uniformly underpredicted chronological age in blood samples

Improved methods for age, frailty, and mortality prediction in large-scale longevity studies · Harvard · PhD

Lost to a baseline

Direct prediction of remaining mouse lifespan using DNA methylation BeadChip data (LOOCV R=0.71) performed worse than metabolomics-based prediction (LOOCV R=0.85)

Improved methods for age, frailty, and mortality prediction in large-scale longevity studies · Harvard · PhD

Tried and failed

biological pathway enrichment on differential chromatin peaks applied to developmental stage-specific open chromatin regions. Outcome: no signal. Reason: small candidate peak sets lacked functional enrichment and appeared to be background noise

Modeling and integration of multi-omics data to study regulatory landscapes governing placenta development · Iowa State · PhD

Considered and rejected

Considered and rejected: Rejected standard library-size normalization alone for mitotic ChIP-seq because global reduction in chromatin-bound factor causes severe underestimation (MAPD 123-145% vs 62-68% for qPCR scaling).

Mining The Spatial And Temporal Context By Which Transcription Factors Occupy Chromatin · Penn · PhD

Tried and failed

subgroup analysis by single measurement technology applied to meta-analysis of DNA methylation biomarkers. Reason: restricting to a single assay technique did not reduce inter-study effect size heterogeneity

Genetic, epigenetic and environmental determinants of cervical cancer risk and implications for early detection and screening · Imperial · PhD

Lost to a baseline

Raw methylation data UMAP embeddings exhibited less defined tissue-specific clusters, stronger batch-specific clustering, and weaker sex-specific separation compared to contextualized MethylGPT sample embeddings.

On the Quantification of Aging · Harvard · PhD

Considered and rejected

Considered and rejected: Rejected selecting differentially methylated probes at a global cross-tissue level in favor of tissue-specific differential methylation analysis due to strong tissue-of-origin signals.

SUPERVISED METHYLATION SIGNATURES FOR TISSUE-SPECIFIC CANCER ETIOLOGICAL FACTORS · Johns Hopkins · Master's

Considered and rejected

Considered and rejected: Rejected strict binary methylation classification (0 or 1) on treated samples because it masks partial (~20%) demethylation changes across the genome.

Elucidating the effects of DNA modification on transcription · Oxford · PhD

Considered and rejected

Considered and rejected: Rejected using weighted methylation for GWAS across accessions because it is distorted by structural/copy number variants, using unweighted mean methylation instead.

Genome - epigenome dynamics and natural variation in Thlaspi arvense · Tübingen · PhD

Tried and failed

unsupervised hierarchical clustering on coverage data applied to single-cell chromatin accessibility profiles. Outcome: no signal. Reason: broad epigenetic state variations confounded underlying genomic copy number variations

Multi-Omics Integration Through Single-Cell Copy Number Analysis In Cancer · Penn · PhD

Tried and failed

subclustering single-nucleus chromatin accessibility data applied to infiltrating immune cell subpopulations. Outcome: data insufficient. Reason: low single-nucleus recovery caused batch effects to overpower biological cell-type separation

SINGLE-NUCLEUS MULTI-OMIC INVESTIGATIONS RESOLVE AND IDENTIFY NOVEL FEATURES OF FIBROLAMELLAR CARCINOMA · Cornell · PhD

Tried and failed

deconvolution of sequence interaction counts applied to repetitive RNA chromatin binding. Outcome: did not generalise. Reason: failed to replicate locus-specific RNA enrichment patterns reported by targeted affinity-purification experiments

Statistical and computational methods for spatial and regulatory genomics · Harvard · PhD

Tried and failed

correlating transposase footprinting activity with proximity-labeled protein abundance applied to measuring transcription factor chromatin binding dynamics. Outcome: no signal. Reason: computational footprint scores did not correlate with physical chromatin-associated transcription factor abundance measured by mass spectrometry

Accelerating the functional prioritization of gene regulatory biomarkers in acute myeloid leukemia · Harvard · PhD

Tried and failed

motif-centered spatial interaction enrichment analysis applied to candidate chromatin architectural transcription factors. Outcome: no signal. Reason: candidate factor motifs showed no enrichment for long-range interactions across tested conditions

Statistical methods for transcription factor footprinting in 3D genome assays · Harvard · PhD

Tried and failed

chromatin accessibility footprinting algorithms applied to transcription factor binding site detection. Outcome: no signal. Reason: controlling for sequencing cut coverage depth eliminates most footprinting signals

Characterizing Regulatory Elements and Non-Coding Variants in the Human Genome · Harvard · PhD

Tried and failed

long-fragment sequencing read analysis applied to chromatin microcompartment resolution. Outcome: no signal. Reason: fragments over 120 bp lost fine resolution and caused microcompartment signals to fade after cohesin depletion

Statistical methods for transcription factor footprinting in 3D genome assays · Harvard · PhD

Considered and rejected

Considered and rejected: Rejected using publicly available Hi-C data to annotate super-enhancer promoter loops because enhancer regulation and chromatin looping dynamics are highly cell-type and chromatin-modifier specific.

The Epigenetic Role of Nuclear Receptor Co-Repressor 2 (NCOR2) in Neuroendocrine Prostate Cancer Progression · University at Buffalo · PhD

Considered and rejected

Considered and rejected: Rejected using sequencing (Hi-C/high-throughput 3C-seq) to read out chromatin conformation because map resolution scaling made detecting small ~15kb regulatory loops cost-prohibitive compared to targeted PCR 3C.

Regulation of neural stem cell behaviors in Drosophila · Harvard · PhD

Tried and failed

conventional bisulfite PCR cloning and sequencing applied to DNA methylation baseline quantification. Reason: cloning and colony-picking bias produced inaccurate methylation estimates

Exploring the function of NLRP7 in regulating human maternal imprinting · Imperial · PhD

Tried and failed

chromatin immunoprecipitation sequencing applied to transcription factor profiling. Outcome: no signal. Reason: samples failed standard quality control metrics

Transcription factor antagonism regulates heterogeneity in embryonic stem cell states · MIT · PhD

Considered and rejected

Considered and rejected: Rejected relying on uncorrected read ends due to hexamer primer-induced methylation bias at the 5' and 3' ends.

Evaluation of Computational and Experimental Parameters in RNA Bisulfite Sequencing Analysis and Applications in Brain Development Studies · Virginia Tech · Doctorate

Tried and failed

enzymatic cleavage assay at low concentrations applied to single-locus chromatin accessibility quantification. Outcome: no signal. Reason: low cleavage probability at discriminative enzyme concentrations led to poor measurement resolution

Principled Methods and Models for Deep Learning Based Functional Genomics · MIT · PhD

Tried and failed

Sonication of high cell density cross-linked chromatin applied to chromatin immunoprecipitation fragmentation. Outcome: no signal. Reason: Excessive cell concentration prevented efficient shearing, leaving fragments over 1 kb and abolishing qPCR enrichment

The role of G-quadruplex DNA in the transcriptional reprogramming of drug-resistant ovarian cancer cells · Imperial · PhD

Tried and failed

chopped tissue sample preparation for ATAC-seq applied to plant seedling chromatin accessibility profiling. Outcome: no signal. Reason: unusually low signal portion of tags score during peak calling

Plastid to Nucleus Signalling and the Evolution of C4 Photosynthesis · Cambridge · PhD

Tried and failed

uniform fixed-size binning instead of peak detection applied to chromatin accessibility feature extraction. Outcome: worse than baseline. Reason: Fixed windows fail to capture variable-width signal boundaries, missing more than half of active regions.

Allelic imbalance of chromatin accessibility in cancer identifies candidate causal risk variants and their mechanisms · Harvard · PhD

Tried and failed

narrow-peak calling on broad epigenetic enrichment domains applied to broad histone modification profiling. Outcome: no signal. Reason: algorithm model assumes narrow enriched regions rather than diffuse, broad heterochromatic enrichment domains

Decoding epigenetic adaptation to adjuvant endocrine therapies in hormone-dependent breast cancer · Imperial · PhD

Considered and rejected

Considered and rejected: Window sizes greater than 600 bp (tested 1000 bp) for chromatin state prediction were rejected due to diminishing performance returns on validation sets while decreasing effective resolution.

From Arabidopsis to Zea: Learning conserved cis mechanisms of gene regulation · Cornell · PhD

Tried and failed

narrow peak calling algorithms applied to broad genomic enrichment domains. Outcome: no signal. Reason: narrow peak callers fail to accurately capture diffuse, broad epigenetic deposition patterns

Epigenome Changes Due to Aging and Reconstitution to a Youthful State Using Targeted and Generic Means · Harvard · Master's

Considered and rejected

Considered and rejected: Peak calling analysis (MACS2 broad peaks) was rejected in favor of full continuous raw alignment profile analysis because epigenetic landscapes act as continuous analog noise histograms rather than discrete peaks.

Epigenome Changes Due to Aging and Reconstitution to a Youthful State Using Targeted and Generic Means · Harvard · Master's

Left open

Perform ChIP-seq on C. elegans mutation accumulation lines to identify histone modifications underlying chromatin-based epimutations and evaluate co-transcriptional establishment. Blocker: Requires wet-lab experimental facilities, C. elegans mutation accumulation lines, and ChIP-seq assays.

Characteristics of spontaneous epimutations · Imperial · PhD

Left open

Probe whether replisomes push cohesin over 20 kb to termination sites using ChIP assays at short intervals without transcription. Blocker: Requires wet lab facilities and yeast strains to perform chromatin immunoprecipitation (ChIP) assays

Replisome components in sister chromatid cohesion establishment · Imperial · PhD

Left open

Identify specific target genes epigenetically regulated by histone lactylation in Plasmodium falciparum using genomic profiling methods like ChIP-seq. Blocker: Requires wet-lab parasite cell culture, chromatin immunoprecipitation (ChIP-seq/CUT&RUN), and sequencing for histone lactylation marks

Histone Lactylation Dynamics in Plasmodium falciparum · Harvard · Master's

Left open

Profile Cited2-chromatin interactions using CUT&RUN or ChIP-seq and measure histone acetylation via CUT&Tag in DRG neurons. Blocker: Requires wet-lab experimental facilities, mouse models, and molecular biology assays (CUT&RUN, ChIP-seq, CUT&Tag)

The role of Cited2 in axonal regeneration · Imperial · PhD

Left open

Determine other histone modifications involved in seed germination and characterize combinatorial epigenetic regulation of gene expression. Blocker: Requires molecular biology wet-lab experiments (e.g., ChIP-seq for various histone marks, plant assays in Arabidopsis).

Functional characterization of histone binding protein ENAP1 in seed germination and drought response · UT Austin · PhD

Left open

Perform ChIP-seq on mouse tissues to directly determine chromatin and histone modifications induced by combined oral contraceptive pill treatment. Blocker: Requires wet-lab facilities, animal tissue samples, and wet-lab ChIP-seq experimental protocols

Molecular mechanisms of ovarian cancer prevention · Imperial · PhD

Left open

Replicate ChIP-seq experiments across 1 to 7 hours post-infection to confirm CTCF binding kinetics in HSV-1 infected chromatin. Blocker: Requires wet lab facilities, HSV-1 viral cultures, and chromatin immunoprecipitation (ChIP-seq) experiments

UNDERSTANDING THE ROLE OF HISTONE H2A VARIANTS IN THE GENOME-WIDE REGULATION OF HSV-1 LYTIC CHROMATIN DYNAMICS · Cornell · PhD

Left open

Add read de-duplication and chromatin accessibility clustering to a scATAC-seq pipeline to separate malignant from non-malignant cells across tumor datasets. Blocker: None

Mutations and disorder in cellular systems · Imperial · PhD

Left open

Quantify the direct causal effects of structural variants on DNA methylation patterns and target gene expression across cancer patient cohorts. Blocker: None

Cancer Regulatory Networks and Susceptibilities: a Computational Framework to Identify Novel Therapeutic Targets in Cancer · Cornell · PhD

Left open

Profile genome-wide DNA methylation across the matched longitudinal patient cohort using EM-Seq or whole-genome bisulfite sequencing. Blocker: Requires access to physical patient tissue samples and a wet lab for DNA methylation sequencing

Identification and characterisation of somatic regulatory mutations in the breast cancer genome · Imperial · PhD