SolvedByAI

Queue

628 cases were harvested and are backed by a real paper, but they fail the test the index publishes on. Each one is listed here with the reason, so the exclusion can be checked rather than trusted.

All 628 assisted 358 not a result 183 unsupported 87
Paper Year Cited Why it is held
Crop Monitoring Using Satellite/UAV Data Fusion and Machine Learning Remote Sensing 2020 305 assisted The machine learning models serve as regression tools inside a human-designed remote-sensing pipeline, and the paper's contribution is an assessment of data-fusion strategies and a comparison of model accuracies for estimating crop traits, not an AI system performing crop monitoring end to end. “The goal of this research was to assess the potential of combining canopy spectral information with canopy structure features for crop monitoring usin...”
Machine learning identifies candidates for drug repurposing in Alzheimer’s d... Nature Communications 2021 293 assisted DRIAD produces a ranked nomination of candidate drugs within a human-run discovery pipeline (human cell-culture perturbations feed it, and its output requires further human validation and trials), so the AI informed rather than solved drug repurposing. “We propose that the DRIAD method can be used to nominate drugs that, after additional validation and identification of relevant pharmacodynamic biomar...”
Matching patients to clinical trials with large language models Nature Communications 2024 283 assisted TrialGPT produces eligibility predictions and rankings that inform human screeners who still complete the recruitment workflow, as evidenced by the user study framing the system as reducing human screening time rather than performing recruitment itself, and its 87.3% criterion-level accuracy (below expert performance) is a prediction evaluated inside a human pipeline. “our user study reveals that TrialGPT can reduce the screening time by 42.6% in patient recruitment”
InteractionGraphNet: A Novel and Efficient Deep Graph Representation Learnin... Journal of Medicinal Chemistry 2021 261 assisted This is a methods paper reporting a new predictive model's benchmark accuracy on existing tasks (affinity prediction, virtual screening, pose scoring) against baselines, functioning as a scoring component for human-run drug design pipelines rather than an AI agent producing a novel end-to-end scientific result. “Extensive binding affinity prediction, large-scale structure-based virtual screening, and pose prediction experiments demonstrated that IGN achieved b...”
A foundation model for clinician-centered drug repurposing Nature Medicine 2024 239 assisted TxGNN produces ranked drug-repurposing predictions validated retrospectively against clinician off-label use, with outputs explicitly framed as candidates for human expert investigation rather than the AI itself establishing any new therapeutic use. “TxGNN's drug-repurposing predictions are accurate, consistent with off-label drug use, and can be investigated by human experts through multi-hop inte...”
AMPlify: attentive deep learning model for discovery of novel antimicrobial... BMC Genomics 2022 221 assisted AMPlify's role was to prioritize candidate peptide sequences, but the actual discovery, confirming antimicrobial activity, was completed by human-run wet-lab bioactivity assays, making this a prediction component inside a larger human pipeline. “demonstrate its utility in prioritizing peptide sequences derived from the Rana [Lithobates] catesbeiana (bullfrog) genome”
MetaRNN: differentiating rare pathogenic and rare benign missense SNVs and I... Genome Medicine 2022 190 assisted The deep learning models are presented as decision-support tools that 'help' humans identify and prioritize variants and whose scores feed downstream human genotype-phenotype analyses, i.e., benchmarked accuracy on an existing prediction task within a human interpretation workflow rather than the AI itself delivering a solved result. “we present pathogenicity prediction models, MetaRNN and MetaRNN-indel, to help identify and prioritize rare nonsynonymous single nucleotide variants (...”
Deep learning suggests that gene expression is encoded in all parts of a co-... Nature Communications 2020 172 assisted The deep learning model predicts mRNA abundance from sequence, but the paper's actual contribution, the discovery of regulatory rules, motif interactions, and co-evolution, was made by humans analyzing the model's outputs within a human-run scientific pipeline, not produced directly by the AI. “By searching for DNA regulatory motifs across the gene regulatory structure, we discover that motif interactions could explain the whole dynamic range...”
Leveraging Uncertainty in Machine Learning Accelerates Biological Discovery... Cell Systems 2020 153 assisted The AI model's uncertainty-guided predictions served to prioritize compounds within a human-run pipeline where the actual discovery (nanomolar kinase affinity, M. tuberculosis growth inhibition) was established by human experimental validation, not by the AI itself. “identifying and experimentally validating compounds with nanomolar affinity for diverse kinases and whole-cell growth inhibition of Mycobacterium tube...”
Prediction of dengue outbreak in Selangor Malaysia using machine learning te... Scientific Reports 2021 142 assisted This is a model-comparison study reporting accuracy metrics for predicting dengue outbreaks, where the predictions are intended to inform health officials who would take action, rather than an AI system end-to-end solving the outbreak problem. “five districts in Selangor, Malaysia, that demonstrated the highest incidence of dengue fever from 2013 to 2017 were evaluated for the best machine le...”
Improving deep learning protein monomer and complex structure prediction usi... Nature Methods 2024 130 assisted DeepMSA2 is an MSA-construction pipeline whose output feeds existing deep learning structure predictors; it improves those predictors' inputs rather than itself performing the structure prediction end to end. “optimization of input information to deep learning-based structure prediction methods must be considered with as much care as the design of the predic...”
Random forest-based prediction of stroke outcome Scientific Reports 2021 130 assisted The abstract reports a Random Forest model's predictive accuracy (AUC) on the existing clinical task of stroke-outcome prognostication, framing the model as a tool that 'can be effectively used' within a human-run cohort study to inform clinical assessment, rather than an AI autonomously producing a standalone result. “machine learning algorithms RF can be effectively used in stroke patients for long-term outcome prediction of mortality and morbidity”
Artificial Intelligence Tool for Optimizing Eligibility Screening for Clinic... JCO Clinical Cancer Informatics 2020 127 assisted The paper evaluates an AI matching tool's agreement, sensitivity, and specificity against manual review within a human-run screening workflow, with the abstract itself framing the output as 'system-assisted' determinations rather than the AI autonomously performing trial screening. “System-assisted trial eligibility determinations were substantially faster than manual review, and the system reliably excluded ineligible patients fo...”
Alzheimer’s disease diagnosis and classification using deep learning techniq... PeerJ Computer Science 2022 122 assisted The abstract merely reports benchmark accuracies of existing CNN architectures (DenseNet-169, ResNet-50) on an Alzheimer's classification dataset, positioning the model as a proposed diagnostic aid for error-prone manual diagnosis rather than an AI that itself solved the clinical task end to end. “The proposed model is usable for real-time analysis and classification of Alzheimer's disease.”
Machine learning-guided discovery of ionic polymer electrolytes for lithium... Nature Communications 2023 115 assisted The ML workflow only recommended candidate ionic liquids; humans then selected, synthesized, and electrochemically tested the actual electrolytes and cells, so the AI prediction was one component inside a human-run experimental pipeline. “By selecting subsets of the recommended ILs, combining with a rigid-rod polyelectrolyte and a lithium salt, we develop a series of thin (~50 μm) and r...”
Machine learning-powered antibiotics phenotypic drug discovery Scientific Reports 2019 111 assisted The machine learning is one component of a human-run phenotypic screening pipeline whose output (MoA predictions) informs human chemists guiding SAR optimization, rather than the AI itself producing the discovery end to end. “machine learning-powered dataset analysis, effectively allowed us to narrow down, compare and predict compound mode of action (MoA)”
Pediatric Severe Sepsis Prediction Using Machine Learning Frontiers in Pediatrics 2019 110 assisted The paper reports a predictive model's retrospective accuracy (AUROC) on an existing clinical prediction task as a decision-support tool to inform clinicians, not an AI end-to-end solving sepsis detection in practice. “the machine learning algorithm achieved an AUROC of 0.916 for discrimination between severe sepsis and control pediatric patients at the time of onset...”
Machine Learning and Deep Learning Approaches in Breast Cancer Survival Pred... Folia Biologica 2019 108 assisted The paper reports comparative accuracies of standard off-the-shelf classifiers on an existing prediction task (breast cancer survival from registry data), with the predictions ultimately serving to inform human treatment-protocol decisions rather than the AI itself constituting the solved contribution. “The results indicated that the multilayer perceptron (MLP), random forest (RF) and decision tree (DT) classifiers could predict survivorship, respecti...”
Machine learning to predict cardiovascular risk International Journal of Clinical Practice 2019 104 assisted The paper evaluates and compares the predictive performance of existing ML methods against risk scales within a human-run cohort analysis, rather than the AI itself producing the end-to-end result as the contribution. “To analyse the predictive capacity of 15 machine learning methods for estimating cardiovascular risk in a cohort and to compare them with other risk s...”
Analysis of missense variants in the human genome reveals widespread gene-sp... The American Journal of Human Genetics 2022 89 assisted The machine learning approach and random-forest MutScore were components wielded by human researchers in a human-driven analysis workflow, and the paper reports a predictor's accuracy on the existing task of variant pathogenicity prediction to aid (not replace) human interpretation of variants of uncertain significance. “We used a machine learning approach to analyze the within-gene distribution of missense variants observed in hereditary conditions and cancer.”
DeepDRK: a deep learning framework for drug repurposing through kernel-based... Briefings in Bioinformatics 2021 85 assisted DeepDRK is a predictive model whose outputs (drug-response scores and repurposing candidates) serve to prioritize drugs for downstream human validation in a precision-medicine workflow, rather than the AI itself completing drug repurposing end to end. “offering an alternative way to prioritize repurposing drugs in precision cancer treatment”
Integrating central nervous system metagenomics and host response for diagno... Nature Communications 2022 82 assisted The machine learning classifier is one component of a combined diagnostic workflow (alongside wet-lab mNGS sequencing run by humans), and the reported result is the accuracy of that integrated assay rather than the AI performing diagnosis end to end. “By leveraging the specificity of mNGS and the sensitivity of an MLC created from CSF host transcriptomes, the combined assay has high sensitivity (88....”
Automated detection of cardiovascular disease by electrocardiogram signal an... Cardiovascular Diagnosis and Therapy 2020 81 assisted The CNN's ECG classifications are benchmarked against human expert labels on an existing diagnostic task, and the paper explicitly frames the model as a decision-support aid whose output informs human clinicians rather than an agent that autonomously completed the clinical diagnostic problem. “The proposed CNN model could help reduce misdiagnosis and missed diagnosis in primary care settings and also improve efficiency and save manpower cost...”
External validation of a deep learning electrocardiogram algorithm to detect... International Journal of Cardiology 2021 80 assisted This is a human-designed and human-run external validation study that reports the accuracy (AUC, sensitivity, specificity) of a previously derived algorithm on an existing detection task, so the AI is the component being evaluated rather than the agent producing the paper's result. “We assessed the performance of the AI-ECG to identify LVSD in this distinct patient population.”
Early Prediction of Mortality, Severity, and Length of Stay in the Intensive... Frontiers in Medicine 2021 75 assisted The ML models produce retrospective risk predictions whose stated role is to inform clinicians' treatment and resource decisions, making the AI a decision-support component in a human-run clinical workflow rather than the agent that solved the clinical problem. “can provide a comprehensive early warning of sepsis, which will contribute to the formulation and management of clinical decisions and the allocation...”
Knowledge-driven drug repurposing using a comprehensive drug knowledge graph Health Informatics Journal 2020 73 assisted The ML system produces repurposing candidate predictions validated retrospectively against known diabetes treatments, with actual repurposing decisions and case-study investigation left to human researchers, making this a prediction component in a larger human workflow. “The evaluation demonstrates that the knowledge-driven approach can produce high predictive results for known diabetes mellitus treatments by only usin...”
On classifying sepsis heterogeneity in the ICU: insight using machine learni... Journal of the American Medical Informatics Association 2019 73 assisted The ML models are experimental instruments in a human-driven study whose contribution is the insight that stratifying sepsis subpopulations improves feature selection, not an AI system solving sepsis prediction end to end. “features selected using sepsis subpopulations as background knowledge yield a superior performance in distinguishing septic from non-septic patients r...”
The Immune Subtypes and Landscape of Gastric Cancer and to Predict Based on... Frontiers in Immunology 2021 71 assisted The paper's primary contribution is the human-driven identification and validation of immune subtypes via clustering on transcriptomic data, with the deep learning model serving only as a downstream component that predicts those human-defined subtypes from pathology images. “Finally, deep learning can predict the immune subtypes well.”
Easy-Prime: a machine learning–based prime editor design tool Genome biology 2021 68 assisted Easy-Prime is a predictive design tool whose ML output (prime editor designs) informs human researchers who then perform and validate the actual genome editing, making it a component in a human-run workflow rather than an AI-produced result. “We provide optimized PE design for installation of 89.5% of 152,351 GWAS variants.”
Optimizing fraud detection in financial transactions with machine learning a... Expert Systems 2024 66 assisted This is a human-run comparative evaluation of six standard ML models on an existing fraud-detection task, where the contribution is performance metrics and insights rather than an AI system end-to-end solving fraud detection. “This paper provides valuable insights into the efficacy of machine learning models for financial fraud detection and emphasizes the importance of stri...”
Mortality Prediction of Septic Patients in the Emergency Department Based on... Journal of Clinical Medicine 2019 65 assisted The deep learning model only produces retrospective mortality risk scores to inform clinicians, with the actual clinical task of identifying and managing septic patients remaining a human-run workflow, as the abstract itself frames the system as assistance rather than an agent that completed the task. “We expect that deep learning can effectively assist medical staff in early identification of critical patients.”
Smart Platform for Data Blood Bank Management: Forecasting Demand in Blood S... Information 2023 65 assisted The AI/ML system is explicitly a decision support tool whose forecasts and classifications inform human-run blood collection, inventory, and scheduling operations rather than autonomously producing the supply-chain outcomes. “We use machine learning and time series forecasting models to develop an AI/ML decision support system.”
Novel antimicrobial peptide discovery using machine learning and biophysical... Drug Development Research 2019 62 assisted The machine-learning prediction was one component of a human-run pipeline: ML generated candidate 20-mers, but humans selected sequences, synthesized them, and performed the wet-lab antimicrobial/cytotoxicity/hemolysis assays that produced the actual discovery result. “Of those, we selected sixteen sequences for synthesis and further study, and evaluated their antimicrobial, cytotoxicity, and hemolytic activities.”
Machine learning-based meta-analysis reveals gut microbiome alterations asso... Nature Communications 2025 61 assisted The ML models were one analytical component inside a human-designed and human-run meta-analysis, reporting classification accuracy on existing microbiome data (without cross-study generalization), while the paper's actual contribution is a human-synthesized overview of PD-microbiome associations rather than an AI-produced end-to-end result. “Within most studies, microbiome-based machine learning models could accurately classify PD patients. However, models were study-specific and did not g...”
Biosensor and machine learning-aided engineering of an amaryllidaceae enzyme Nature Communications 2024 56 assisted The AI (MutComputeX) generated predicted enzyme variants that were then screened, validated, and completed via a human-run biosensor/wet-lab pipeline, so the improved biocatalyst is the product of the workflow rather than the AI alone. “A structure-based residual neural network (MutComputeX) is subsequently developed and used to generate activity-enriched variants of a plant methyltra...”
A hybrid deep learning framework for early detection of diabetic retinopathy... Scientific Reports 2025 54 assisted The paper's contribution is a human-designed hybrid CNN-RNN architecture demonstrated via benchmark classification accuracy on existing public datasets against other models, so the AI's predictions serve to evaluate a method within the human model-development workflow rather than constituting an end-to-end solved capability that is itself the paper's primary result. “Proposed model experimental evaluation demonstrate the superior performance over traditional deep learning models like CNN, RNN, InceptionV3, VGG19 an...”
Machine learning-based ABA treatment recommendation and personalization for... Brain Informatics 2022 51 assisted The ML models only generate treatment-goal recommendations that are benchmarked against and intended to inform clinician-prepared ABA plans, with the actual therapy still delivered by human practitioners in a clinical workflow. “The patient similarity and collaborative filtering methods predicted ABA treatment with an average accuracy of 81–84%, with a normalized discounted cu...”
Prediction of incident cardiovascular events using machine learning and CMR... European Radiology 2022 47 assisted This is a human-run cohort study reporting accuracy/AUC of an SVM classifier on the well-established task of cardiovascular risk prediction, with humans performing cohort selection, segmentation, and feature extraction, so the ML is a predictive component evaluating radiomics' incremental value rather than an agent that solved a problem. “To assess the model performance, the accuracy, sensitivity, specificity, and AUC were reported.”
A machine learning model trained on a high-throughput antibacterial screen i... PLoS Computational Biology 2022 47 assisted The ML model serves as a compound-prioritization step within a human-run drug discovery pipeline, where researchers experimentally tested the top-ranked predictions in wet-lab growth inhibition assays to obtain the actual hits. “the developed ML approach can be used for compound prioritization before screening, increasing the typical hit rate of drug discovery”
Machine learning for optimized individual survival prediction in resectable... Journal of Cancer Research and Clinical Oncology 2022 45 assisted The paper reports comparative accuracy of ML models on the existing task of survival prediction and delivers a human-built prognostic risk-score tool intended to inform surgical oncologists' clinical judgment, making the AI a component in a human-run clinical workflow rather than a standalone AI-produced result. “we could establish a compact risk score model with 20 input parameters and thus provide a clinical tool to improve prediction of oncological outcome a...”