Chapter Four · failure evidence
What High-Throughput Assays & Screening got wrong, from 75 dissertations
High-throughput screening platforms and computational pipelines face substantial reproducibility barriers when primary hits fail during orthogonal secondary testing. Experimental scaling is further constrained by analytical throughput bottlenecks, cellular measurement artifacts, physiological disconnects, and intractable computational costs. These records come from PhD theses at 20 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.
Analytical latency and labor-intensive protocols bottleneck screening throughput
Standard analytical separation assays, manual reading steps, and multi-step protocols create operational backlogs that make high-throughput scaling impractical. Switching settling delays in microfluidic electronics, high reagent expenses, and low-efficiency filter conjugation further limit candidate library evaluation.
Considered and rejected
Considered and rejected: Rejected the HCS γH2AX assay as a high-throughput primary screening tool for food contact materials due to lengthy sample preparation and low throughput.
Evaluation of the suitability of in vitro bioassays for the genotoxicity assessment of food contact materials Entwicklung und Optimierung von bio-assays zur Bestimmung der Genotoxizität von Lebensmittelkontaktmaterialien · DSpace-CRIS at TU Wien
Lost to a baseline
High-throughput 96-well cellulose acetate filter conjugation had ~5-fold lower efficiency (4.7 x 10^-7 transconjugants/donor) than standard 50-mL analytical filter conjugation.
Synthetic biology approaches for engineering bacteria as living therapeutics · MIT
Considered and rejected
Considered and rejected: Single VH-primer amplification was used initially for rapid candidate screening, rejecting immediate dual-chain amplification across all single cells due to low throughput.
Isolation and characterization of the functional breadth of neutralizing antibodies that cross-react with diverse sarbecovirus strains · ResearchWorks
Considered and rejected
Considered and rejected: Rejected untargeted screening (HPLC-ICP-MS analysis of all individual isolates) due to excessive analytical instrument time and lack of throughput.
High-throughput isolation of anaerobic arsenic-transforming microorganisms · EPFL
Considered and rejected
Considered and rejected: Directed evolution random-mutagenesis screening on ssnBVMO was halted after Round 3 due to high NADPH costs, large-scale purification limits, and low plate-screening throughput, pivoting instead to rational engineering.
Investigation of Thermostable Enzymes Identified Through the Use of Sequence Similarity Networks · Queens University Institutional Repository
Considered and rejected
Considered and rejected: Rejected direct high-throughput screening on electrode arrays for whole-genome search because direct electrochemical testing across thousands of mutants was impractical.
ENGINEERING BIOLOGICAL SYSTEMS TO ADVANCE ELECTROMICROBIAL PRODUCTION · Cornell
Considered and rejected
Considered and rejected: Rejected direct gene targeting via homologous recombination because screening ~42,000 transformants to find a targeted insertion was impractical.
Promoter analysis in transgenic sugar beet · De Montfort Open Research Archive (DORA)
Considered and rejected
Considered and rejected: Rejected standard PAGE/HPLC/MS assays as primary high-throughput screening tools due to being cumbersome, denaturing, and low-throughput, developing the fluorescent G-quadruplex sensor instead.
Investigating the Mechanism of Non-Enzymatic RNA Replication · Harvard
Tried and failed
radioactivity- or conductivity-based chromatography and TLC applied to high-throughput lipid quantitative relative profiling. Outcome: too slow. Reason: unsuited for high-throughput quantitative profiling compared to modified ELISA and mass spectrometry methods
A lipidomic approach to elucidating the effects of antifungal compounds on signalling in fungi · Imperial
Tried and failed
manual expert triage in high-throughput screening applied to large-scale disease screening programmes. Outcome: too slow. Reason: expert reader bottlenecks caused massive operational backlogs, leaving most collected images unanalyzed
Modelling active case-finding strategies for tuberculosis in urban India · Imperial
Tried and failed
Time-multiplexed switching between sensing and actuation electrodes applied to High-throughput flow-through single-cell processing. Outcome: too slow. Reason: Switching transients between high-voltage actuation and sensing electronics caused millisecond measurement settling delays limiting throughput
Single-Cell Impedance Cytometry for the Control of Single-Cell Electroporation in a Flow-Through Device · Georgia Tech
Considered and rejected
Considered and rejected: Smart-seq2 without UMIs for high-throughput single-cell experiments (suffered from low throughput of ~few hundred cells per plate, PCR amplification bias, and inefficient template switching for lowly expressed transcripts)
Considered and rejected
Considered and rejected: Rejected inkjet printing because it prints only a single reagent at a time and multi-step protocols limit high-throughput scalability
Considered and rejected
Considered and rejected: Novel enclosed quartz evaporation boat rejected for industrial high-throughput scaling due to slow evaporation rates despite superior film uniformity
CHARGE TRANSPORT MEASUREMENTS ON AMORPHOUS SELENIUM PHOTORECEPTOR FILMS · HARVEST
Primary screening hits fail to reproduce during secondary or orthogonal validation
Apparent hits identified in primary phenotypic, genetic, or bioassay screens frequently fail to validate in secondary assays due to assay artifacts and normalization errors. Screened candidates across functional genomics and directed evolution libraries show poor replication across independent replicates or orthogonal testing platforms.
Tried and failed
high-throughput phenotypic screening for candidate target selection applied to therapeutic target validation in cancer cells. Outcome: did not generalise. Reason: top-scoring screening candidate failed to reproduce apoptotic or anti-proliferative effects in secondary validation assays
Modulation of miR-361-3p alters apoptosis in endocrine-resistant and -responsive breast cancer · Imperial
Tried and failed
high-throughput directed evolution screening applied to engineered enzyme variants. Outcome: did not generalise. Reason: apparent screening hits were false positives caused by copurifying contaminants affecting initial normalization and readout
Tried and failed
genome-wide pooled screening with targeted validation applied to chemotherapy drug sensitisation targets. Outcome: did not generalise. Reason: hit depletion scores failed to reproduce during targeted follow-up screens
CRISPRi screens to identify combination therapies for the improved treatment of ovarian cancer · MIT
Tried and failed
mass spectrometry screening followed by western blot validation applied to identifying viral-host protein degradation targets. Outcome: no signal. Reason: candidate targets identified by proteomics failed to show differential expression on orthogonal validation
ZER1 CONTRIBUTES TO THE GROWTH OF HPV-POSITIVE CANCERS · Penn
Tried and failed
Forward genetic mutagenesis screening applied to identifying pathogen-resistant mutant lines. Outcome: did not generalise. Reason: Primary screen candidates failed to demonstrate statistically significant survival increases upon secondary validation assays
Tried and failed
High-throughput phenotyping screen using automated behavioral tracking applied to Metagenomic bacterial library bioactivity screening. Outcome: did not generalise. Reason: Variations in bacterial food lawn density produced non-reproducible behavioral artifacts across confirmation screens.
Tried and failed
arrayed CRISPR knockout genetic screening applied to identifying viral host dependency factors. Outcome: did not generalise. Reason: CRISPR screen hits failed to replicate results from a prior genome-wide siRNA screen
Tried and failed
pooled CRISPR loss-of-function screening across independent knockdown conditions applied to synthetic lethality target identification in cancer. Outcome: did not generalise. Reason: Enrichment detected in only one shRNA condition arm failed to replicate across independent knockdown replicates.
SMARCB1 Maintains Lineage Fidelity in Clear Cell Renal Cell Carcinoma · Cambridge
Tried and failed
bioassay screening of microbial isolates applied to antimicrobial and toxicity hit identification. Outcome: did not generalise. Reason: initial phenotypic bioactivity hits failed to reproduce during subsequent follow-up testing
An exploration of chemical cues as mediators of marine predator-prey interactions · Georgia Tech
Tried and failed
yeast surface display kinase activity screening applied to drug-resistant kinase mutant identification. Outcome: did not generalise. Reason: enrichment in yeast display screen did not replicate in orthogonal biochemical and cellular validation assays
Tried and failed
nanoliter-scale combinatorial library synthesis and screening applied to protein-protein interaction inhibitor discovery. Outcome: no signal. Reason: Screening crude reaction mixtures yielded no reproducible hits; initial signals were assay artifacts.
Generating macrocyclic inhibitors of protein-protein interactions · EPFL
Tried and failed
fixed cell microarray screening for protein interactions applied to validating viral glycoprotein host receptors. Outcome: no signal. Reason: Microarray binding hits failed to reproduce when assayed on live cells via flow cytometry.
Structural basis of antibody-mediated immunity against Crimean-Congo hemorrhagic fever virus glycoproteins · UT Austin
Tried and failed
phenotypic drug screening with morphological linear models applied to 3D organoid size perturbations. Outcome: did not generalise. Reason: Screening hits predicting size increases failed to replicate in 3D validation experiments
Assay artifacts, cell handling losses, and poor sensitivity degrade measurement fidelity
Plate artifacts, low fluorescent sensitivity, and excessive cell loss during sorting compromise the detection of true hits in high-throughput workflows. Biological donor variability, off-target viability phenotypes, and reagent contamination further introduce false positives and mask subtle cellular interactions.
Tried and failed
high-throughput screening with primary immune cells applied to co-culture functional genomics screens. Outcome: unstable. Reason: post-expansion cellular impurity and high donor-to-donor variability compromised assay reproducibility
Considered and rejected
Considered and rejected: Rejected relying solely on siRNA screening hits for target validation due to extensive seed-mediated off-target effects and false-positive viability phenotypes.
The Role of RUVBL1/RUVBL2 and Their Potential as Therapeutic Targets in Non-Small Cell Lung Cancer · DSpace at UTSWMED
Considered and rejected
Considered and rejected: Discarded exclusive reliance on single-protein commercial line blots due to high false-positivity rates compared to tissue indirect immunofluorescence assay screening.
Characterising Autoimmune Neurologic Disorders: Biomarker Identification and Clinical Phenotyping · Research Repository UCD
Considered and rejected
Considered and rejected: Rejected using the direct spTorA-mTurquoise2 periplasmic translocation assay for high-throughput Tat inhibitor screening due to insufficient fluorescent assay sensitivity.
PQS-dependent quorum sensing in pseudomonas aeruginosa is linked to protein export via the twin-arginine translocation (tat) system · University of Nottingham Repository
Lost to a baseline
Soluble TCR kinetic screening on Octet RED384 was less sensitive than the cellular Jurkat-NFAT activation assay for detecting weak peptide interactions (e.g., failed to fit 57 PSL peptides).
Strategies and methods for the evaluation of T cell receptor cross-reactivity in cancer immunotherapeutic approaches · Publikationssystem UB Tuebingen
Considered and rejected
Considered and rejected: High-throughput forward genetics screening in standard mismatch repair-defective HCT116 cells without an inducible mutagenesis switch; rejected due to high background mutations, low clone yields, and need for subjective microscopy to distinguish clonal vs non-clonal growth
Combining Chemical Biology and Forward Genetics to Identify the Target of Anticancer Small Molecules · DSpace at UTSWMED
Considered and rejected
Considered and rejected: RNA-based and FRET-based biosensors were rejected for high-throughput screening due to narrow ligand classes, poor in vivo translation, and low signal-to-noise ratios.
Measuring chemicals with biology : engineering genetic biosensors for chemical analysis · UT Austin
Tried and failed
ATP-based luminescence assay for cell viability applied to drug-perturbed high-throughput cell screening. Reason: Severe plate artifacts and poor correlation with cell counts due to drug-altered ATP generation and morphology
A Framework for The Study of Compound Interactions in L1000 · Harvard
Lost to a baseline
High-throughput flow cytometry failed to detect weaker increases in phagocytic success over time for 3% acrylamide gels that were clearly detected by manual confocal microscopy
Biofilm viscoelasticity and microstructures impinge on immune clearance · UT Austin
Considered and rejected
Considered and rejected: Rejected using pour-plating for colony PCR screening due to agar contamination inhibiting PCR and causing false positives.
Establishment of tools for genetic modification of the thermophilic methanogenic archaeon Methanothermobacter thermautotrophicus deltaH · Publikationssystem UB Tuebingen
Tried and failed
cell-pooling in single-cell droplet screening applied to single-cell knockout perturbation screens. Outcome: worse than baseline. Reason: signal dilution across unperturbed cells within overloaded droplets reduced out-of-sample accuracy
Investigating human disease mechanisms through population genetics and experimental genetic screens · Harvard
Considered and rejected
Considered and rejected: Rejected standard FACS isolation directly on low-yield MFD setups due to high cell-loss rates (~70% loss during sorting), necessitating redesign of high-capacity radial microfluidics or optical screening.
Virtual screening selects candidates with poor drug properties, insolubility, and cytotoxicity
Computationally docked candidates frequently fail in experimental validation because top-scoring compounds exhibit mammalian cytotoxicity and poor drug-likeness. In silico screening hits also suffer from poor solubility, lack of binding affinity in biophysical assays, or structural network instability.
Tried and failed
structure-based virtual screening and molecular dynamics refinement applied to enzyme inhibitor discovery. Reason: hit compound had poor drug-like properties with high molecular weight, high lipophilicity, and mammalian cell cytotoxicity
Structure-based discovery of lipoteichoic acid synthase inhibitors. · Imperial
Tried and failed
ensemble molecular docking and virtual screening applied to enzyme inhibitor discovery. Reason: hit compound had poor drug-likeness (high molecular weight, high clogP) and mammalian cytotoxicity
Structure-Based Discovery of Lipoteichoic Acid Synthase Inhibitors. · Cambridge
Tried and failed
Structure-based virtual screening with ensemble docking applied to bacterial enzyme inhibitor discovery. Reason: Top-scoring hit exhibited mammalian cytotoxicity and poor drug-likeness (high molecular weight and lipophilicity)
Structure-Based Discovery of Lipoteichoic Acid Synthase Inhibitors. · Cambridge
Tried and failed
structure-based virtual screening for pharmacological chaperones applied to destabilized mutant protein rescue. Outcome: no signal. Reason: most computationally docked candidates failed to rescue protein levels in cellular validation assays
Disease Mechanisms and Therapeutic Strategies in Munc18-1 Encephalopathies · Cornell
Tried and failed
host-directed small molecule phenotypic screening applied to blood-stage antiparasitic drug discovery. Reason: host-targeted candidate compounds induced severe host cell toxicity and eryptosis, precluding translational utility
Tried and failed
pharmacophore linking virtual screening applied to enzyme active site ligand discovery. Outcome: no signal. Reason: Screened compounds lacked binding affinity in crystallographic soaking and thermal shift assays due to poor solubility.
Characterization and targeting of the 2-methylcitrate cycle in Pseudomonas aeruginosa · Cambridge
Tried and failed
similarity-based virtual screening of monomers applied to thermoset polymer synthesis. Outcome: unstable. Reason: Screened comonomer candidate failed to yield a stable crosslinked network during experimental synthesis
Designing Macromolecules using Machine Learning and Simulations · MIT
Tried and failed
polymer-lipid hybrid nanoparticle formulation applied to in vivo mRNA delivery screening. Reason: formulations caused severe in vivo systemic toxicity requiring exclusion
In vitro and surrogate host screens fail to reflect physiological in vivo performance
Binding motifs and enzyme activity identified in vitro or in heterologous hosts fail to translate to robust activity in living organisms. Screening phenotypes in simplified culture systems or non-equilibrium synthesis methods often bias selection toward artifacts that do not match physiological conditions.
Tried and failed
in vitro high-throughput motif enrichment screening applied to identifying in vivo protein-RNA binding motifs. Outcome: did not generalise. Reason: weakly enriched in vitro motifs failed to show corresponding binding peaks in in vivo crosslinking validation
Tried and failed
in vitro high-throughput secondary motif identification applied to in vivo RNA-protein binding enrichment. Outcome: did not generalise. Reason: secondary low-affinity binding motifs detected in vitro lacked robust enrichment in in vivo crosslinking-immunoprecipitation peaks
Tried and failed
early-timepoint screening to predict delayed fractional outgrowth applied to pathogen detection in low-moisture matrices. Outcome: did not generalise. Reason: Early pre-screening failed to correlate with or predict extended-incubation fractional positive recovery for slow-growing strains.
Difficult-to-detect Salmonella strains and novel Listeria species - implications for industry · Cornell
Considered and rejected
Considered and rejected: Decided against whole-cell screening in deep-well plates due to poor correlation with cell-free extract activity.
Biocatalytic synthesis of chiral amine building blocks · University of Nottingham Repository
Tried and failed
in vitro enzyme activity screening in heterologous host applied to candidate enzyme selection. Outcome: no signal. Reason: Screening in E. coli produced indistinguishable activity levels across candidates, failing to predict in vivo performance.
Bioproduction of L-piperazic acid in gram scale using Aureobasidium melanogenum · Imperial
Tried and failed
organoid drug screening across culture conditions applied to tumor cell state-specific drug sensitivity. Outcome: no signal. Reason: standard ex vivo culture conditions failed to produce consistent, discernable differential sensitivities across isogenic states
Cell states and therapeutic targets in pancreatic cancer · Harvard
Tried and failed
single-cell FACS screening for secreted small molecules applied to strain engineering for metabolite overproduction. Outcome: worse than baseline. Reason: biases selection toward intracellular retention rather than secretion, lowering total yield
Enhancing secreted production from yeast species with high throughput microdroplet screening methods · UT Austin
Considered and rejected
Considered and rejected: Rejected thin film sputtering method for high-throughput combinatorial alloy screening because non-equilibrium synthesis produces phases inconsistent with bulk phase diagrams and hinders mass-production scaling.
Investigation on Electrodeposition of Metals and Alloys with Advanced Characterizations · Georgia Tech
Statistical filtering and sample pooling strategies cause signal loss or inflated errors
Naive hypothesis screening and rigid p-value cutoffs lead to premature filtering of genuine treatment effects in small samples. Pooling strategies with dilution effects and unconfirmed suspect workflows either inflate follow-up testing burdens or penalize true positives through overly conservative corrections.
Tried and failed
sample splitting with hard p-value threshold screening applied to multiple testing outcome selection. Outcome: worse than baseline. Reason: premature filtering of true signals due to low sample size or weak treatment effect thresholds
MODERN ADVANCES IN ACCOUNTING FOR UNMEASURED CONFOUNDING VIA SENSITIVITY ANALYSIS · Penn
Lost to a baseline
Naive hypothesis screening performs worse than full-sample Bonferroni correction in small samples, fewer candidate hypotheses, or low Gamma_con due to premature filtering of signals.
MODERN ADVANCES IN ACCOUNTING FOR UNMEASURED CONFOUNDING VIA SENSITIVITY ANALYSIS · Penn
Considered and rejected
Considered and rejected: Rejected using a single universal suspect screening workflow/filter threshold without standard confirmation due to high false detection rates in in silico matching and compound loss across technical replicates.
Assessing the Global Occurrence, Fate, and Ecotoxicological Effects of Contaminants of Emerging Concern in the Environment Using Target Analysis and Suspect Screening · DSpace at SUNY Buffalo
Considered and rejected
Considered and rejected: Rejected FWER-controlling multiple-testing corrections due to overly conservative true-positive penalties in high-throughput data.
Tried and failed
linear model peak calling across sliding windows applied to cross-assay enhancer activity detection. Outcome: did not generalise. Reason: peak calls showed poor consistency across different high-throughput reporter assay protocols
Tried and failed
quality-based sample screening prior to testing applied to high-dimensional hypothesis testing without reference data. Outcome: no signal. Reason: Screening does not improve test performance without an external reference database.
High-dimensional statistical models and hypothesis tests with a focus on forensic and genetic applications · Iowa State
Tried and failed
correlated group testing with dilution effects applied to hierarchical specimen screening. Outcome: worse than baseline. Reason: detecting multiple positives per pool increases follow-up testing, reducing screened individuals per test relative to naive pooling
Tackling COVID-19 Challenges at Cornell University: Stochastic Modeling, Simulation, and Statistics · Cornell
High computational scaling costs make exhaustive virtual screening intractable
Direct density functional theory and non-linear programming methods scale poorly with candidate library size and constraint volume, forcing researchers to reject high-fidelity computational screening. Simplified single-target or closed-shell docking approximations miss key reaction surface physics and multitarget polypharmacology effects.
Considered and rejected
Considered and rejected: Restricting high-throughput virtual screening exclusively to closed-shell/singlet reaction coordinates was rejected for open-shell transition metal catalysis because it misclassifies ground-state spin surfaces and misses distortion-tuning effects.
Considered and rejected
Considered and rejected: Rejected TD-DFT excited-state geometry relaxation/adiabatic energy calculations for virtual screening due to prohibitive computational expense
High-throughput virtual screening of molecules for photon conversion · Imperial
Considered and rejected
Considered and rejected: Rejected superstructure MINLP targeting methods for large-scale screening due to poor polynomial/exponential constraint scaling
Considered and rejected
Considered and rejected: Rejected direct high-fidelity DFT redox calculation for large-scale screening due to prohibitive computational costs on uncharacterized candidate libraries.
Atomistic Modeling and Machine Learning for the Rational Design of Organic Energy Storage Materials · Georgia Tech
Considered and rejected
Considered and rejected: Single-target docking virtual screening models rejected in favor of proteome-scale multitarget interaction signatures to account for polypharmacology and evade mutation-driven resistance.
Rigorous Evaluation of Comparability and Fidelity of Drug Repurposing Technologies · DSpace at SUNY Buffalo
Left open by the authors
Problems the authors named and did not get to.
Left open
Experimentally evaluate synthesized factorizable mAb libraries via high-throughput screening panning against targets. Blocker: Requires a wet lab and physical reagents for high-throughput screening of mAb libraries
Safe and Ethical Implementation of Intelligent Systems · MIT
Left open
Identify and validate targets for the category 1 and category 3 screening hits from the ΔugtP pathway-directed screen. Blocker: Requires wet-lab experimental biology and biochemistry facilities to validate molecular targets of small molecule hits
Discovery of highly selective inhibitors to probe the physiology of the bacterial cell envelope · Harvard
Left open
Construct in vivo RiPP combinatorial libraries for high-throughput activity-based screening against therapeutic protein-protein interaction targets. Blocker: Requires wet lab facilities, biological reagents, and experimental screening infrastructure
Design of Post-Translationally Modified Peptides by Combining Enzymes from Diverse Pathways · MIT
Left open
Perform high-throughput proteomics screening and train a deep learning model on surface electronics data to predict protein surface azidation. Blocker: Requires wet-lab high-throughput proteomics screening experiments to generate training data.
Azide-Containing Hypervalent Iodine Reagents for Biomolecules Functionalization · EPFL
Left open
Perform high-throughput knockout screens and apply machine learning models to identify proteins involved in cholesterol trafficking. Blocker: Requires wet-lab biological facilities to execute high-throughput knockout screens
EXPLORATION AND DEVELOPMENT OF XYLOPYRINDE FLUORESCENT TOOLS FOR SPATIOTEMPORAL IMAGING · Penn
Left open
Perform library-based screening across thousands of targets and train machine learning models to predict the optimal PE6 variant for given target sites. Blocker: Requires a wet lab to perform library-based high-throughput screening across thousands of target sites to generate training data.
Enhancing precision genome editing agents through directed evolution and protein engineering · Harvard
Left open
Perform large-scale screens to establish guide design and target site selection rules for RADARS sensors targeting lowly expressed transcripts. Blocker: Requires a wet lab to perform large-scale experimental screening and molecular assays in biological cells.
Base Editing–Enabled Technologies and Multiplex Genome Editing · Harvard
Left open
Apply neural networks and high-throughput screening to discover next-generation photothermal materials and simulate solar desalination architectures. Blocker: Lack of specific target properties, dataset specifications, or concrete simulation model details
Design Architectures and Analysis of Solar Driven Evaporative Technology for Desalination · DSpace at SUNY Buffalo
Left open
Formally validate multi-targeted screening hits as bona fide pan-DUB inhibitors using biochemical assays. Blocker: Requires physical screening hit compounds and wet-lab biochemical validation assays
Discovery and Development of Novel Deubiquitinase Inhibitors via Parallel High-Throughput Screening · Harvard
Left open
Develop an active learning model to predict optimal reporter cell lines for phenotypic drug screening a priori. Blocker: Requires wet-lab experimental validation and additional cell line profiling data beyond the 93 clones tested
Single-Cell Image Analysis Enables High-Throughput Phenotypic Drug Screen and Elucidates Cell-Fate Decision Principles · DSpace at UTSWMED
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.