What Biomarker Discovery & Profiling got wrong, from 39 theses

Tried and failed

apoptotic and cell cycle biomarker expression profiling applied to predicting radiotherapy tumour response. Outcome: no signal. Reason: biomarker expression levels failed to correlate significantly with tumour regression grade across evaluated datasets

A next-generation -omics analysis of the radiotherapy response in rectal cancer · Imperial · PhD

Tried and failed

immunofluorescence biomarker quantification on tissue microarrays applied to breast cancer clinical prognostic stratification. Outcome: no signal. Reason: No statistically significant correlation was found between histone mark intensity and subtype markers or patient survival

COMPREHENSIVE FUNCTIONAL AND MOLECULAR PROFILING AND INTEGRATIVE ANALYSES OF PRIMARY AND METASTATIC TRIPLE NEGATIVE BREAST CANCERS · Harvard · Master's

Tried and failed

evaluating metabolic gene expression as prognostic biomarker applied to cancer patient survival datasets. Outcome: no signal. Reason: gene expression levels did not statistically correlate with overall or disease-free survival

Delineating the role of malate dehydrogenase 2 in the hypoxia-driven metabolic reprogramming of metastatic cancers · University College Dublin · PhD

Tried and failed

statistical association of standard clinical prognostic biomarkers applied to disease progression and treatment response. Outcome: no signal. Reason: standard clinical variables lacked statistically significant correlation with minimal residual disease and post-induction clinical outcomes

High-efficiency multi-drug functional profiling in a microfluidic device towards personalized predictions in pediatric leukemia · Georgia Tech · PhD

Tried and failed

Correlation analysis for predictive biomarker discovery applied to synthetic lethality across cancer cell lines. Outcome: no signal. Reason: No robust lineage-agnostic or lineage-specific biomarkers or transcription factors correlated with dual dependency.

Systematic Interrogation of CBP/p300 Dependency in Cancer · Harvard · PhD

Tried and failed

statistical correlation analysis of circulating biomarkers applied to elevated disease plasma proteins. Outcome: no signal. Reason: biomarkers lacked statistically significant correlation despite all showing elevated expression levels

The impact of multiple myeloma and carfilzomib on the endothelium · Imperial · PhD

Tried and failed

1H NMR metabolomic profiling with multivariate analysis applied to longitudinal urinary biomarkers for physiological state. Outcome: no signal. Reason: Unsupervised and supervised multivariate models showed no predictive metabolic shift (CV-ANOVA p>0.05, Q2Y<0)

Immunology of choriodecidua and onset of labour · Imperial · PhD

Tried and failed

biomarker association regression analysis applied to metabolic and blood pressure outcomes. Outcome: no signal. Reason: circulating biomarker levels showed no significant correlation with glycemic control or blood pressure measures

VITAMIN B12, ONE-CARBON METABOLISM, AND METABOLIC HEALTH IN WOMEN OF REPRODUCTIVE AGE · Cornell · PhD

Tried and failed

tertile-based biomarker stratification in regression models applied to longitudinal cognitive decline prediction. Outcome: no signal. Reason: Total baseline cerebrospinal biomarker levels did not differentiate categorical cognitive impairment outcomes over time

The association of cerebrospinal fluid α-synuclein, level of olfactory disability, and enteric nervous system dysfunction with motor disability, cognitive deficits, and anxiety in · Harvard · Master's

Tried and failed

baseline biomarker profiling to predict intervention outcomes applied to phenotypic response to dietary intervention. Outcome: no signal. Reason: baseline clinical, metabolic, and endocrine characteristics showed no statistically significant correlation with phenotypic improvement

Longitudinal Assessments of Ovarian Morphology in Women with PCOS: Implications for Diagnosis and Hypocaloric Interventions · Cornell · PhD

Tried and failed

multivariable linear regression of biomarker variance applied to predicting clinical health disorder prevalence. Outcome: no signal. Reason: biomarker mean and variability showed no statistically significant association with disease incidence

Associations between close-up dry period and postpartum health disorders for dry cows fed a negative dietary cation-anion close-up diet · Cornell · Master's

Tried and failed

baseline biomarker quantification via sandwich ELISA applied to longitudinal neurodegenerative motor progression. Outcome: no signal. Reason: Total baseline biomarker levels showed no meaningful correlation with longitudinal disease score progression across 36 months.

The association of cerebrospinal fluid α-synuclein, level of olfactory disability, and enteric nervous system dysfunction with motor disability, cognitive deficits, and anxiety in · Harvard · Master's

Tried and failed

candidate biomarker panel screening in biofluids applied to disease risk discrimination in saliva cohort. Outcome: no signal. Reason: candidate microRNA markers failed to show statistically significant differences or discriminative ability (AUC 0.5-0.6) in validation

A COMPARATIVE SPECIES APPROACH TO UNDERSTAND THE ROLE OF MIRNAS IN MAMMARY GLAND HEALTH AND DISEASE · Cornell · PhD

Tried and failed

nanopore barcoded probe multiplexed detection applied to circulating miRNA cancer biomarkers in serum. Outcome: no signal. Reason: multiplex panel showed no significant difference between healthy control and cancer serum samples

Single-molecule detection of biomarkers for hepatocellular carcinoma and cholangiocarcinoma using novel nanopore-based sensing platforms · Imperial · PhD

Tried and failed

stratified biomarker survival analysis via step-function binarization applied to multi-omics cancer drug response data. Outcome: did not generalise. Reason: biomarker survival association directions contradicted established literature across different cohorts and treatments

Informatics Approaches for Identifying Drug-Specific Markers and Deciphering Genetic Regulation Mechanism in Cancer Treatment · Georgia Tech · PhD

Tried and failed

functional genetic screens for biomarker identification applied to cancer chemotherapy response prediction. Outcome: did not generalise. Reason: in vitro functional screen hits did not overlap with published clinical multi-gene prognostic signatures

CRISPRi screens to identify combination therapies for the improved treatment of ovarian cancer · MIT · PhD

Tried and failed

proteomic biomarker association analysis applied to cardiovascular disease progression. Outcome: did not generalise. Reason: associations were directionally inconsistent across different clinical endpoints and life stages

Valvular Heart Disease Stages Among Older Adults And The Proteomic Markers of Aortic Stenosis: The Atherosclerosis Risk in Communities Study · Harvard · Master's

Tried and failed

Multivariate metabolic profiling for biomarker identification applied to personalized dietary intake assessment. Outcome: did not generalise. Reason: Identified biomarkers lacked specificity across diverse demographics, leading to contradictory and uninterpretable dietary feedback

Developing a Personalised Nutrition toolkit for the nutritional management of individuals at risk of cardiovascular disease · Imperial · Master's

Tried and failed

supervised multivariate regression on metabolic profiles applied to dietary assessment in free-living individuals. Outcome: did not generalise. Reason: controlled trial biomarkers had poor agreement and systematic bias when applied to free-living populations

Applying metabolic profiling as an objective dietary assessment method for personalised nutrition · Imperial · PhD

Tried and failed

automated vessel segmentation biomarker tracking across cohorts applied to longitudinal clinical magnetic resonance angiography. Outcome: did not generalise. Reason: vessel count reduction over time observed in one clinical trial cohort failed to reproduce in another

Automated precision computational image analysis and applied machine learning for experimental and clinical hematology applications · Georgia Tech · PhD

Tried and failed

evaluating published transcriptomic biomarker signatures directly applied to independent paediatric disease diagnosis. Outcome: did not generalise. Reason: original models failed to meet required clinical diagnostic performance thresholds on independent paediatric cohort

Applying multi-omics analyses to investigate Mycobacterium tuberculosis infection: a dual perspective · Imperial · PhD

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 · PhD

Tried and failed

adding clinical features to logistic regression applied to cancer risk prediction. Outcome: worse than baseline. Reason: additional biomarkers and demographic predictors provided no incremental discrimination over primary biomarker and age

The Utility of CA125 for the Detection of Ovarian Cancer in Primary Care · Cambridge · PhD

Tried and failed

combining multiple correlated biomarkers in survival forest applied to disease progression prediction. Outcome: no signal. Reason: multiplexing biomarkers provided no statistically significant predictive gain over the strongest standalone marker

Predicting Progression from Mild Cognitive Impairment to Alzheimer’s Disease using Blood-Based Biomarkers and Amyloid-Beta PET: A Survival Analysis Approach · Harvard · Master's

Tried and failed

combining engineered spatial biomarkers with foundation model embeddings applied to survival prediction from histopathology images. Outcome: no signal. Reason: biomarker features provided no additional predictive signal beyond what foundation models already captured

Artificial Intelligence-Based Phenotyping of the Tumor Microenvironment in Hematoxylin and Eosin-Stained Images of Solid Tumors · Georgia Tech · PhD

Tried and failed

combining multi-platform assay features for classification applied to protein biomarker diagnostic signatures. Outcome: worse than baseline. Reason: supplementary platform measurements did not provide additional discriminative value beyond single-platform markers

Using host “omic” datasets to better understand and diagnose paediatric infectious and inflammatory diseases · Imperial · PhD

Lost to a baseline

Automated CVS biomarkers (AUC 0.81-0.87) did not replicate the perfect discrimination (100% sensitivity/specificity) previously achieved by manual expert ratings of all lesions using 40% or 50% cutoffs.

Statistical Techniques For Addressing The Clinico-Radiological Paradox In Multiple Sclerosis · Penn · PhD

Lost to a baseline

For predicting adverse in-hospital events during AHF, all novel admission biomarkers had lower discrimination than baseline serum creatinine (AUC 0.65).

Validation of novel biomarkers of Acute Kidney Injury · University College Dublin · PhD

Tried and failed

correlating circulating biomarker levels with tissue expression applied to colorectal cancer biomarker validation. Outcome: no signal. Reason: circulating plasma concentrations did not correlate with immunohistochemical tissue expression levels

Expression pattern and clinicopathological relevance of neurotensin and its receptors in colorectal cancer · Imperial · PhD

Tried and failed

urinary multiplex biomarker profiling applied to disease severity stratification and monitoring. Outcome: unstable. Reason: High intra- and inter-subject biomarker variability obscured meaningful clinical signals

Clinical monitoring and biomarkers to stratify severity and predict outcomes in children with cystic fibrosis: CLIMB-CF · Imperial · PhD

Considered and rejected

Considered and rejected: Rejected primary biomarker discovery in urine due to low proteomic coverage when total urine protein concentration is low

Maladaptive Repair in Acute Kidney Injury · Johns Hopkins · PhD

Tried and failed

RT-qPCR validation of sequencing biomarker candidates applied to low-abundance biofluid microRNA biomarkers. Outcome: no signal. Reason: Target microRNAs were undetectable in more than half of the biofluid samples

MicroRNAs in extracellular vesicles as biomarkers for pancreaticobiliary cancers · Imperial · PhD

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 · PhD

Tried and failed

multivariable regression for prognostic biomarker identification applied to cancer genomic mutations. Outcome: no signal. Reason: Univariate biomarker associations lost statistical significance due to co-occurrence with confounding covariates.

Development of Methods for Cancer Genome Analysis and Clinical Applications · Harvard · PhD

Tried and failed

mass-spectrometry metabolomic profiling with multiple testing correction applied to cancer risk biomarker discovery. Outcome: no signal. Reason: No individual metabolites remained statistically significant after adjusting for multiple hypothesis testing.

Exploring Pre- and Postmenopausal Breast Cancer Risk Factors through Metabolomics and Risk Prediction Modeling · Harvard · PhD

Considered and rejected

Considered and rejected: Rejected assessing high-order interactions among 15 biomarkers and baseline covariates due to multiple testing issues and model overspecification risk.

Measuring the impact of reduced antibiotic use in hospital settings using electronic health records · Oxford · PhD

Tried and failed

stability selection with LASSO logistic regression applied to high-dimensional proteomic biomarker discovery. Outcome: no signal. Reason: No individual proteomic features met stability selection thresholds after adjusting for clinical covariates.

Molecular phenotyping of severe asthma using statistical and machine learning models · Imperial · PhD

Tried and failed

regularised regression biomarker profiling applied to disease risk prediction from metabolomics. Outcome: no signal. Reason: derived dietary metabolomic signatures showed no statistically significant association with disease risk in multivariable models

Diet, Metabolomics, and Colorectal Cancer Risk · Harvard · PhD

Considered and rejected

Considered and rejected: Rejected soluble full-length Ataxin-3 in PBMCs as a standalone prognostic disease biomarker due to weak, non-significant correlations with disease progression scores (SARA, INAS, CSDP) compared to mutant Ataxin-3 and NfL.

Parkin und Ataxin-3 als potenzielle Biomarker in der Spinozerebellären Ataxie Typ 3 (SCA3) · Tübingen · PhD

Tried and failed

univariate biomarker classification applied to necrotizing enterocolitis diagnosis. Outcome: no signal. Reason: individual pathway metabolites lacked sufficient discriminatory power as standalone diagnostic markers

Mapping the metabolome in the developing gut · Imperial · PhD

Tried and failed

predictive regression using single metabolomic biomarkers applied to perinatal depression prediction. Outcome: no signal. Reason: individual metabolite levels showed weak correlations and lacked statistical predictive power

Phenotyping perinatal depression : exploring interactions among biopsychosocial and behavioral determinants · UT Austin · PhD

Left open

Validate phase angle as a prognostic stratification biomarker in extensive longitudinal cohorts across different cancer types. Blocker: Requires collecting longitudinal clinical and bioimpedance data from large cancer patient cohorts

Impacto de un programa de ejercicio físico en el ángulo de fase y su correlación con la salud física y mental en pacientes de cáncer durante y después de la pandemia de COVID-19 · University of Deusto · PhD

Left open

Identify clinical biomarkers predicting patient response to low-dose SFK inhibition combined with BRAF/MEK blockade. Blocker: Requires clinical patient cohorts, tissue samples, or wet-lab experimental validation.

Tumor cell-intrinsic signals promoting tolerance and adaptation to oncogenic kinase inhibition · MIT · PhD

Left open

Develop predictive biomarkers for patient tumor sensitivity to AHR or kynurenine pathway inhibition. Blocker: Requires clinical trial cohorts, patient tumor tissue samples, and wet lab experimental validation

Novel Functions of the Transcription Factor Aryl Hydrocarbon Receptor (AHR) and Its Tryptophan-Derived Ligands in Cancer Cells · UT Southwestern · PhD

Left open

Validate candidate molecular, genomic, and epigenetic biomarkers from mucosal tissue, serum, or stool for colorectal cancer risk stratification in commercial clinical practice. Blocker: Requires clinical patient biospecimens (tissue, serum, stool) and wet lab diagnostic infrastructure.

Optimising decision-making in the management of dysplasia in inflammatory bowel disease · Imperial · PhD

Left open

Validate lipoprotein-cholesterol biomarkers and test lipidomics/genomics profiles across larger, stratified cohorts of cancer patients. Blocker: Requires wet lab facilities, biological patient cohorts/samples, and clinical lipidomic/genomic profiling infrastructure

Protein Engineering for Personalized therapy and diagnosis · EPFL · PhD

Left open

Establish quantitative score thresholds in the Biomarker Toolkit to predict clinical translation success using historical biomarker evaluation datasets. Blocker: None

Development and validation of the biomarker toolkit: a tool aiming to quantifiably assess biomarker utility and guide development. · Imperial · PhD

Left open

Validate the 34 lipid signatures associated with medulloblastoma metastasis across human patient cohorts. Blocker: Requires access to human patient cohorts and clinical wet-lab mass spectrometry facilities to validate lipid biomarkers.

Applications of High-resolution Mass Spectrometry and Matrix-assisted Laser Desorption/Ionization Mass Spectrometry Imaging-based Non-targeted Metabolomics in Biomarker Discovery · Georgia Tech · PhD

Left open

Apply the Biomarker Toolkit scoring checklist to published non-cancer biomarker studies to validate rationale-related evaluation attributes across disease categories. Blocker: None

Development and validation of the biomarker toolkit: a tool aiming to quantifiably assess biomarker utility and guide development. · Imperial · PhD

Left open

Investigate synaptic changes, inflammation, and mitochondrial dysfunction in prodromal Parkinson's disease cognitive deficits using biological or longitudinal assays. Blocker: Requires wet lab biological assays, biomarker data, or specialized longitudinal cohort samples measuring synaptic/mitochondrial markers

Prodromal Parkinson’s Disease and its Implications for Longitudinal Studies · Harvard · PhD

Left open

Analyze Parkinson's Disease Biomarkers Program (PDBP) proteomic data across cohorts to evaluate biological implications of molecular clusters. Blocker: PDBP multi-omics clinical and proteomic data requires controlled-access data use agreements and approval.

Investigating Molecularly Defined Clusters of Parkinson’s Disease Based on Multi-Omics Data With Clinical and Biological Implications · Harvard · Master's