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
Proteomics and Machine Learning Approaches Reveal a Set of Prognostic Marker... Frontiers in Physiology 2021 76 assisted The ML (SVM classifier) was one component in a human-run proteomics workflow, humans generated the mass spectrometry data, selected/validated the biomarkers via MRM assays, and performed the drug screening, so the AI only informed the human-driven discovery rather than producing the result itself. “we employed supervised machine learning (ML) approaches using a linear support vector machine model to identify the classifiers of patients with non-s...”
Forecasting of COVID-19 epidemic size in four high hitting nations (USA, Bra... Applied Computing and Informatics 2020 57 assisted The authors ran a human-driven forecasting study and merely employed an off-the-shelf tool (Fb-Prophet) as one component, reporting its projection accuracy on an existing epidemic-forecasting task rather than the AI itself contributing a new end-to-end capability. “the authors adopted a machine learning (ML) model called Fb-Prophet”
Deep Learning-Based Decision-Tree Classifier for COVID-19 Diagnosis From Che... Frontiers in Medicine 2020 287 assisted The deep learning classifier reports accuracy on an existing image-classification task and is explicitly positioned as a pre-screening/triage aid within a human clinical workflow, with RT-PCR remaining the definitive diagnostic test. “The proposed deep learning-based decision-tree classifier may be used in pre-screening patients to conduct triage and fast-track decision making befor...”
Detection and analysis of COVID-19 in medical images using deep learning tec... Scientific Reports 2021 158 assisted The paper reports classification accuracies of pre-trained CNN models on existing public datasets and explicitly frames the AI output as an aid to human clinicians rather than an end-to-end solved diagnostic capability. “our work can help virologists and radiologists to make a better and faster diagnosis in the struggle against the outbreak of COVID-19”
An Early Warning Tool for Predicting Mortality Risk of COVID-19 Patients Usi... Cognitive Computation 2021 114 assisted The machine learning (XGBoost) served as a biomarker-identification and risk-scoring component inside a human-run retrospective clinical study, with the resulting nomogram explicitly framed as a tool to inform doctors' patient-stratification decisions rather than the AI itself performing the clinical task end to end. “The prognostic model, nomogram, and LNLCA score can help in early detection of high mortality risk of COVID-19 patients, which will help doctors to im...”
MGP-AttTCN: An interpretable machine learning model for the prediction of se... PLoS ONE 2021 56 assisted The abstract reports a human-designed model's benchmark accuracy on an existing task (early sepsis prediction on MIMIC-III), framing the AI as a predictor whose outputs inform clinicians and contribute to an ongoing research effort rather than the AI itself producing the paper's primary result. “when predicting sepsis five hours prior to onset on our new realistic labels, our proposed model achieves an area under the ROC curve of 0.660 and an...”
Artificial intelligence in sepsis early prediction and diagnosis using unstr... Nature Communications 2021 412 assisted The abstract reports a prediction model's accuracy (AUC 0.94) on the existing task of sepsis early warning and frames the AI's output as decision support to augment physician detection rather than autonomously completing the clinical task of diagnosing and treating sepsis. “We compare the SERA algorithm against physician predictions and show the algorithm's potential to increase the early detection of sepsis by up to 32%...”
Deep-learning-enabled antibiotic discovery through molecular de-extinction Nature Biomedical Engineering 2024 189 assisted The deep-learning models only predicted candidate antimicrobial peptides, and the actual antibiotic discovery was completed by a human-run pipeline of synthesis, wet-lab validation, and mouse infection studies. “We synthesized 69 peptides and experimentally confirmed their activity against bacterial pathogens.”
Risk prediction of cardiovascular disease using machine learning classifiers Open Medicine 2022 158 assisted This is a standard model-accuracy study applying existing classifiers (MLP, K-NN) to a public UCI dataset, reporting benchmark metrics rather than the AI itself solving CVD detection end to end in practice. “Experimental-based results demonstrate that a higher accuracy in detection of 82.47% and an area-under-the-curve value of 86.41% are obtained using th...”
Machine learning applied on chest x-ray can aid in the diagnosis of COVID-19... European Radiology Experimental 2021 145 assisted The CNN ensemble is developed and evaluated as a decision-support tool to aid human COVID-19 diagnosis from CXR (RT-PCR remains the reference standard and the tool is still 'in training'), so the AI informs diagnosticians rather than autonomously solving the diagnostic task. “We aimed to train and test a deep learning classifier to support the diagnosis of coronavirus disease 2019 (COVID-19) using chest x-ray (CXR)”
CheXaid: deep learning assistance for physician diagnosis of tuberculosis us... npj Digital Medicine 2020 136 assisted The paper's primary contribution is a study of whether AI assistance improves clinician diagnostic accuracy; the clinicians made the final diagnoses, with the AI serving as a decision-support aid rather than performing the task end to end as the paper's contribution. “Use of the algorithm resulted in a modest but statistically significant improvement in clinician accuracy”
MRI Deep Learning-Based Solution for Alzheimer’s Disease Prediction Journal of Personalized Medicine 2021 131 assisted The paper reports a deep-learning classifier's benchmark accuracy on an existing task (OASIS MRI diagnosis) that remains a human clinical decision, and its own conclusion frames the system as a tool for computer-assisted rather than autonomous diagnosis. “deep learning-based strategies are an effective tool for building a robust solution for Alzheimer's-assisted diagnosis based on MRI data”
Prediction of survival and metastasis in breast cancer patients using machin... Clinical Epidemiology and Global Health 2018 116 assisted This is a human-run retrospective comparison of off-the-shelf classifiers' prediction accuracy on existing patient records, where the AI models are evaluated components rather than the source of a solved clinical problem. “This study aimed to compare the performance of six machine learning techniques two traditional methods for the prediction of BC survival and metastasi...”
An efficient mixture of deep and machine learning models for COVID-19 diagno... PLoS ONE 2020 112 assisted The AI is explicitly positioned as a computer-aided diagnosis tool providing auxiliary suggestions to human experts who make the actual diagnosis, rather than autonomously solving the diagnostic task. “can provide experts with auxiliary diagnosis suggestions, which can reduce the burden of experts to a certain extent”
Pre-existing and machine learning-based models for cardiovascular risk predi... Scientific Reports 2021 85 assisted The ML models only produce risk scores that inform human clinical decision-making within an existing prediction task, reporting improved accuracy metrics rather than the AI itself solving the clinical problem end to end. “The model can be easily adopted for risk assessment and clinical decision making.”
An Intelligent Early Flood Forecasting and Prediction Leveraging Machine and... Processes 2023 74 assisted The abstract only reports accuracy metrics (MAPE, classification accuracy) of ML models predicting precipitation on an existing public weather dataset, which is a prediction component feeding a human-run flood-warning pipeline rather than an end-to-end solved flood forecasting/alert result. “We employ the dataset from the Australian commonwealth office of meteorology named Historical Daily Weather dataset to test the effectiveness of the p...”
Comparison of risk models for mortality and cardiovascular events between ma... PLoS ONE 2019 69 assisted Machine learning served as an analytical tool inside a human-designed cohort study (Shinken Database) where researchers selected the cohort, outcomes, and comparison with logistic regression, and the abstract merely reports the developed models' AUCs on a standard risk-prediction task rather than the AI itself accomplishing the task as the paper's contribution. “The areas under the curve of the developed models for each event were 0.900, 0.912, 0.879, 0.758, and 0.753, respectively.”
Deep Learning-driven research for drug discovery: Tackling Malaria PLoS Computational Biology 2020 63 assisted The deep learning models performed prediction/virtual screening to prioritize compounds, but the actual discovery was completed by human-run wet-lab experimental evaluation and hit confirmation, making the AI one component in a human pipeline. “The top computational predictions were evaluated experimentally against asexual blood stages of both sensitive and multi-drug-resistant Plasmodium fal...”
Early Prediction of Sepsis Based on Machine Learning Algorithm Computational Intelligence and Neuroscience 2021 58 assisted This is a human-run methodological study reporting accuracy (AUC) of off-the-shelf algorithms on the existing task of early sepsis prediction, with the contribution being human-designed feature-generation and imputation methods rather than an AI system that itself solved the problem. “XGBoost and LightGBM algorithms are both excellent in prediction performance (AUC: 0.910∼0.979)”
Deep Learning for SARS COV-2 Genome Sequences IEEE Access 2021 56 assisted The abstract reports a proposed CNN-Bi-LSTM model's classification accuracy on genome-sequence tasks, positioning deep learning as a tool ('alternative avenue') to inform virus and motif detection rather than the AI itself producing a definitive solved result. “These results are encouraging enough to recognise deep learning algorithms as alternative avenues for detecting SARS CoV-2 as well as detecting regula...”
Matrix Factorization-based Technique for Drug Repurposing Predictions IEEE Journal of Biomedical and Health Informatics 2020 56 assisted The paper's contribution is an enhanced prediction method whose outputs are putative drug-target/disease candidates evaluated via retrospective accuracy metrics and literature checks, i.e., a prediction component feeding a human-driven drug-repurposing pipeline rather than AI-produced, validated repurposing results. “obtaining very good performance, with an Average Precision Score of 0.82”
Predicting High-Risk Prostate Cancer Using Machine Learning Methods Data 2019 55 assisted This is a human-run retrospective evaluation of standard ML algorithms' accuracy on an existing clinical prediction task, where the model's risk predictions serve as decision support for the human screening workflow rather than being themselves the AI-produced result. “We evaluated the accuracy of various machine learning algorithms in predicting high-risk prostate cancer.”
Identification of Novel Antibacterials Using Machine Learning Techniques Frontiers in Pharmacology 2019 47 assisted The ML models only scored/prioritized candidate molecules, and the actual identification and confirmation of antibacterials was completed by human-run wet-lab validation (biological studies, MIC/CC50, in vitro and in vivo inhibitor identification), making AI one component in a human pipeline. “The antibacterial activity of the selected molecules against E. coli was assessed using a comprehensive biological study.”
Deep and Transfer Learning Approaches for Automated Early Detection of Monke... ACS Omega 2023 47 assisted The paper reports benchmark-style accuracy of existing architectures on a public skin-lesion dataset, framing the models as aids that 'could help' human mpox diagnosis rather than the AI itself completing the diagnostic task end to end. “It was concluded that CNN, VGG19, and Inception v3 could help in early detection of mpox skin lesions, and Inception v3 returned the best (96.56%) cla...”
Diagnostic and prognostic capabilities of a biomarker and EMR‐based machine... Clinical and Translational Science 2021 44 assisted The paper reports the accuracy of a predictive scoring model (AUROC 0.83) as a clinical decision-support capability within a human-run diagnostic workflow, not an AI system that itself performs or completes sepsis recognition and management end to end. “The area under the receiver operating curve (AUROC) for diagnosis of sepsis was 0.83.”
Using Machine Learning to Predict the Antibacterial Activity of Ruthenium Co... Angewandte Chemie International Edition 2023 44 assisted The ML models only prioritized candidate compounds within a human-run pipeline: humans synthesized the 54 new complexes and performed the wet-lab antibacterial assays that produced the improved hit-rate, so the AI informed rather than produced the discovery. “The models perform well and are used to predict the activity of 54 new compounds. These displayed a 5.7x higher hit-rate (53.7%) against methicillin-r...”
A Deep Analysis of Transfer Learning Based Breast Cancer Detection Using His... International Conference on Signal Processing and Integrated Networks (SPIN) 2023 44 assisted This is a human-run comparative evaluation of existing pre-trained models on an existing histopathology dataset, reporting accuracy metrics rather than contributing a new end-to-end capability produced by the AI. “we have analyzed pre-trained deep transfer learning models such as ResNet50, ResNetl01, VGG16, and VGG19 for detecting breast cancer using the 2453 hi...”
Using Machine Learning to Predict Synergistic Antimalarial Compound Combinat... Frontiers in Pharmacology 2018 42 assisted CoSynE's predictions served as candidate prioritization within a human-run wet-lab screening pipeline, the synergistic combinations were established by experimental validation, not by the model itself, and the abstract frames the AI's value as reducing experimental effort by 41%. “CoSynE predicted 20 combinations to be synergistic, which was experimentally validated for nine of them (45%), corresponding to an enrichment of 1.70...”
A New Approach to Drug Repurposing with Two-Stage Prediction, Machine Learni... OMICS A Journal of Integrative Biology 2022 40 assisted The ML pipeline only generated candidate drug lists whose efficacy is predicted but not demonstrated, and the abstract explicitly frames the output as informing future human-driven drug discovery and trials rather than solving repurposing itself. “These were previously noted to be effective against two of the three diseases, and have a high probability of being effective against the other.”
Classification and deep-learning–based prediction of Alzheimer disease subty... Translational Psychiatry 2023 37 assisted The deep neural network is a prediction model reporting accuracy within a larger human-driven GWAS study whose primary contributions (subtype identification and kidney-function insight) came from human analysis, not the AI. “We developed a prediction model for LOAD subtypes using a deep neural network, which achieved an accuracy of 0.694 (2870/4137) in the discovery cohort...”
Machine Learning Models for Early Prediction of Sepsis on Large Healthcare D... Electronics 2022 34 assisted This is an experimental benchmark study evaluating ML models' predictive accuracy against existing scoring systems on a retrospective public dataset (MIMIC-III), not an AI system solving sepsis detection end to end. “This work presents an experimental study of some machine-learning-based models for sepsis prediction”
Machine Learning Algorithm Identifies an Antibiotic Vocabulary for Permeatin... Journal of Chemical Information and Modeling 2020 34 assisted The ML algorithm identified candidate fragments, but the result was produced through a human-run pipeline combining theory, computation, and wet-lab experiments to validate the fragments and confirm permeation of selected compounds, making the AI one component rather than the agent that solved the problem. “By synergizing theory, computation, and experiment, we validate our approach, explain the molecular mechanism behind identified fragments promoting co...”
The Large Scale Machine Learning in an Artificial Society: Prediction of the... Computational Intelligence and Neuroscience 2015 26 assisted Machine learning is only one component used to optimize individual behaviors within a human-built artificial society simulation, while the epidemic prediction and intervention conclusions come from the human-designed computational experiment pipeline. “artificial Beijing is reconstructed based on geodemographics and machine learning is involved to optimize individuals' behaviors”
Machine Learning Models for Cardiovascular Disease Events Prediction Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society 2022 22 assisted The paper reports a comparative accuracy benchmark of standard ML models on an existing cohort prediction task, with the outputs intended to inform a future human-run clinical risk-estimation pipeline (TIMELY) rather than the AI itself producing the solved clinical outcome. “These results will be used in the TIMELY study to estimate the risk score and mortality of CVD in patients with 10-year risk.”
Diabetes Prediction Through Linkage of Causal Discovery and Inference Model... Biomedicines 2025 10 assisted The paper applies standard ML models to an existing public Kaggle dataset and reports their accuracy alongside causal-inference analysis, making AI one component in a human-run analytical workflow rather than the producer of a novel end-to-end result. “The study achieved high accuracy across models using logistic regression (84.84%) and deep learning (84.83%).”
Large language models encode clinical knowledge Nature 2023 3,955 assisted This is a benchmark/evaluation study where Flan-PaLM's accuracy on existing medical QA datasets is measured within a human-run evaluation framework, and the authors themselves conclude the best model remains inferior to clinicians, so the clinical task is not solved by the AI. “The resulting model, Med-PaLM, performs encouragingly, but remains inferior to clinicians.”
CADD-Splice—improving genome-wide variant effect prediction using deep learn... Genome Medicine 2021 637 assisted The deep-learning splice scores are pre-existing component predictions that the authors themselves compared, selected, and integrated into their own CADD pipeline, with the humans running the integration and evaluation rather than the AI producing the end result. “We integrate two specialized splicing scores into CADD (Combined Annotation Dependent Depletion; cadd.gs.washington.edu), a widely used tool for genom...”
Efficient evolution of human antibodies from general protein language models Nature Biotechnology 2023 449 assisted The language model only suggested mutations that guided a human-run wet-lab pipeline of screening and laboratory evolution, which is what actually produced the improved antibodies. “We performed language-model-guided affinity maturation of seven antibodies, screening 20 or fewer variants of each antibody across only two rounds of...”
Almanac — Retrieval-Augmented Language Models for Clinical Medicine NEJM AI 2024 401 assisted The AI generates responses that a human expert panel evaluates in a human-run benchmark workflow, and the paper itself frames the system as having 'potential' to aid clinicians in decision-making rather than producing an AI-solved clinical result. “Our results show the potential for LLMs with access to domain-specific corpora to be effective in clinical decision-making.”
Improving de novo protein binder design with deep learning Nature Communications 2023 396 assisted The deep learning models (AlphaFold2/RoseTTAFold as in-silico design assessors, ProteinMPNN as a sequence designer) are components augmenting a pre-existing, human-run energy-based design pipeline that humans completed with wet-lab validation, rather than an AI agent end-to-end producing the binders. “Here, we explore the augmentation of energy-based protein binder design using deep learning.”