Machine Learning Research Articles

Bridging surgical oncology and personalized medicine: the role of artificial intelligence and machine learning in thoracic surgery

Bridging surgical oncology and personalized medicine: the role of artificial intelligence and machine learning in thoracic surgery

Machine learning-assisted investigation on nonlinear vibration analysis of bio-inspired auxetic tubes

PurposeAuxetic tubular structures with negative Poisson’s ratios have gained significant attention in biomedical applications, particularly in vascular and esophageal stents, due to their potential to reduce embolism risks. This study aims to investigate the nonlinear vibration characteristics of such structures and develop accurate predictive models using machine learning (ML) techniques.Design/methodology/approachThe governing equations of auxetic tubes are derived using Hamilton’s principle and von-Kármán’s nonlinear assumptions, while Malek-Gibson relations determine their effective mechanical properties. The mechanical behavior of polylactic acid (PLA) is experimentally analyzed through tensile testing and digital image correlation (DIC) with additional insights from scanning electron microscopy. The nonlinear vibration equations are solved via the Ritz method, and vibrational behavior is assessed using the direct displacement control approach. Predictive modeling is performed using six ML algorithms – CatBoost, decision tree, random forest, gradient boosting tree, extreme gradient boosting (XGBoost) and support vector regression (SVR) – along with an artificial neural network (ANN). Response surface methodology is employed to optimize the effects of edge supports, radius ratios and auxetic cell geometry on vibrational behavior.FindingsThe results demonstrate a strong agreement between ML/ANN predictions and the analytical Ritz method, confirming the reliability of the developed models. The analysis reveals that variations in edge supports, radius ratios and auxetic cell geometry significantly influence the vibrational response of the structures. The optimized configurations enhance the structural performance, making these auxetic tubular metastructures highly suitable for biomedical applications.Originality/valueThis study uniquely integrates analytical modeling, experimental analysis and ML-based predictive modeling to comprehensively assess and optimize the vibrational behavior of auxetic tubular metastructures. The findings provide valuable insights for the design of next-generation auxetic stents, improving their mechanical performance and expanding their potential biomedical applications.

Machine Learning-Based Prediction Models for Healthcare Outcomes in Patients Participating in Cardiac Rehabilitation: A Systematic Review.

Cardiac rehabilitation (CR) has been proven to reduce mortality and morbidity in patients with cardiovascular disease. Machine learning (ML) techniques are increasingly used to predict healthcare outcomes in various fields of medicine including CR. This systemic review aims to perform critical appraisal of existing ML-based prognosis predictive model within CR and identify key research gaps in this area. A systematic literature search was conducted in Scopus, PubMed, Web of Science, and Google Scholar from the inception of each database to January 28, 2024. The data extracted included clinical features, predicted outcomes, model development, and validation as well as model performance metrics. Included studies underwent quality assessments using the IJMEDI and Prediction Model Risk of Bias Assessment Tool checklist. A total of 22 ML-based clinical models from 7 studies across multiple phases of CR were included. Most models were developed using smaller patient cohorts from 41 to 227, with one exception involving 2280 patients. The prediction objectives ranged from patient intention to initiate CR to graduate from outpatient CR along with interval physiological and psychological progression in CR. The best-performing ML models reported area under the receiver operating characteristics curve between 0.82 and 0.91, with sensitivity from 0.77 to 0.95, indicating good prediction capabilities. However, none of them underwent calibration or external validation. Most studies raised concerns about bias. Readiness of these models for implementation into practice is questionable. External validation of existing models and development of new models with robust methodology based on larger populations and targeting diverse clinical outcomes in CR are needed.

AI integration in EHR systems in developing countries: a systematic literature review using the TCCM framework

Purpose This study aims to examine the adoption of artificial intelligence (AI) in electronic health record (EHR) systems in developing countries. It also aims to identify key challenges, explore the effectiveness of AI-driven solutions and propose a structured roadmap for optimizing AI integration to enhance health-care information management. Design/methodology/approach A systematic literature review (SLR) was conducted using the theory-context-characteristics-methodology (TCCM) framework to synthesize existing research on AI integration in EHR systems. The study analyzed peer-reviewed journal articles, conference proceedings and academic reports published between 2019 and 2024. The review applied a multi-stage synthesis process, categorizing studies based on theoretical perspectives, contextual barriers, AI technology characteristics and research methodologies. Findings The study identifies several barriers to AI integration in EHR systems in developing countries, including infrastructural limitations, interoperability challenges, workforce shortages, data fragmentation and regulatory gaps. Despite these challenges, AI-powered EHR solutions, such as machine learning, natural language processing (NLP), predictive analytics and computer vision, demonstrate potential for improving health-care workflows, reducing medical errors and enhancing real-time decision-making. The findings emphasize the need for regulatory frameworks, AI literacy programs, public–private partnerships (PPPs) and open-source AI models to support sustainable AI integration. Originality/value This study contributes to the literature by systematically reviewing AI integration in EHR systems within resource-limited settings, an area that remains underexplored. Unlike previous studies focused on AI implementation in high-income countries, this research provides a context-specific analysis tailored to developing countries. The study introduces a structured roadmap for AI-driven health-care transformation, offering policy recommendations and practical insights for health-care providers, AI developers and policymakers. The findings support digital health strategies aimed at enhancing patient outcomes and optimizing health-care information management in low-resource environments.

Cell Painting and Machine Learning Distinguish Fibroblasts From Nonfailing and Failing Human Hearts

Cell Painting and Machine Learning Distinguish Fibroblasts From Nonfailing and Failing Human Hearts

White matter microglia morphological changes with aging in guinea pig offspring born growth restricted.

Fetal growth restriction is implicated in the programming of later-life neurodegeneration. We hypothesized that growth-restricted offspring would show accelerated changes to microglial white matter morphology, relative to controls. Control guinea pig sows were fed ad libitum, while maternal nutrient restriction sows received 70% of control diet switched to 90% from mid-gestation. Offspring were sacrificed at ∼26 days (neonate) or ∼110 days (adult) postpartum. Coronal brain sections from the frontal cortex were subject to IBA1 staining for microglial detection and analyzed by machine learning software. At birth, total body weight of growth-restricted offspring was reduced relative to control (p<0.0001) with postnatal catch-up growth observed. Microglial density was reduced in the corpus callosum of control (p<0.05) and growth-restricted (p=0.13) adults, relative to neonates. Adults from both groups showed greater IBA1-positive area in the cingulum and periventricular white matter (p<0.05) and increased microglial fractal dimension in the corpus callosum (p<0.10) and periventricular white matter (p<0.05), relative to neonates. At the time points studied, we report age-related changes in white matter microglial morphology. However, maternal nutrient restriction leading to fetal growth restriction in guinea pigs does not appear to exacerbate these white matter microglia morphological changes as a marker for later-life neurodegeneration.

A Machine Learning Approach to Predict Cognitive Decline in Alzheimer Disease Clinical Trials.

Among the participants of Alzheimer disease (AD) treatment trials, 40% do not show cognitive decline over 80 weeks of follow-up. Identifying and excluding these individuals can increase power to detect treatment effects. We aimed to develop machine learning-based predictive models to identify persons unlikely to show decline on placebo treatment over 80 weeks. We used the data from the placebo arm of EXPEDITION3 AD clinical trial and a subpopulation from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Participants in the EXPEDITION3 trial were patients with mild dementia and biomarker evidence of amyloid burden. For this study, participants were identified as those who demonstrated clinically meaningful cognitive decline (CMCD) or cognitively stable (CS) at final visit of the trial (week 80). Machine learning-based classifiers were trained to classify participants into CMCD vs CS groups using combinations of demographics, APOE genotype, neuropsychological tests, and biomarkers (volumetric MRI). The results were developed in 70% of the EXPEDITION3 placebo sample using 5-fold cross-validation. Trained models were then used to classify the participants in an internal validation sample and an external matched sample ADNIAD. Eight hundred ninety-four of the 1,072 participants in the placebo arm of the EXPEDITION3 trial had necessary follow-up data, who were on average aged 72.7 (±7.7) years and 59% female. 55.8% of those participants showed CMCD (∼2 years younger than those without) at the final visit. In the independent validation sample within the EXPEDITION3 data, all the models showed high sensitivity and modest specificity. Positive predictive values (PPVs) of models were at least 11% higher than base prevalence of CMCD observed at the end of the trial. The subset of matched ADNI participants (ADNIAD, N = 105) were aged 74.5 (±6.4) years and 46% female. The models that were validated in ADNIAD also showed high sensitivity, modest specificity, and PPVs of at least 15% higher than the base prevalence in ADNIAD. Our results indicate that predictive models have the potential to improve the design of AD trials through selective inclusion and exclusion criteria based on expected cognitive decline. Such predictive models need further validation across data from different AD clinical trials.

A testbed for landslide prediction with blockchain-based transmission and cloud offloading

Purpose The purpose of this paper is to develop an IoT-based testbed for land displacement monitoring in real-time with blockchain-enabled transmission and machine learning for predictions. Cloud offloading has also been incorporated into the system proposed. Design/methodology/approach The system consists of a modelled landslide testbed at a laboratory scale with soil, water level, humidity, temperature sensors and a designed extensometer using a 3D printer. An Arduino microcontroller handles all sensor information, and a Raspberry Pi performs blockchain and transmission to a gateway using a 4G transmission module. The transmitted data is received in a server GUI application and a webpage where long short-term memory (LSTM) and multi-layer perceptron (MLP) are used for predictions. For further scalability of this system, cloud offloading was implemented, allowing the information to be accessed across multiple platforms. Findings The accuracy of the designed extensometer was compared to an industrial-grade extensometer. Moreover, the predictions performed with MLP and LSTM yielded a MAPE of 6.0% and 7.8%, respectively. Finally, the blockchain analysis demonstrated that using smaller block sizes provides better security but lower throughput than large block sizes. Originality/value A sophisticated testbed for landslide monitoring, which includes blockchain, AI and cloud offloading, has been proposed along with in-depth performance analysis.

Abstract 7434: Cloud-based machine learning for enhanced tumor classification in cancer genomics: an end-to-end solution for whole slide imaging data

Abstract The digitization of histological slides into high-resolution Whole Slide Images (WSIs) has revolutionized pathology workflows, providing unparalleled insights into tissue morphology and playing a vital role in advancing cancer diagnostics and treatment. Tasks that were once time-consuming and subjective can now be automated and enhanced with machine learning (ML) and artificial intelligence (AI) solutions. These include image analysis systems capable of extracting diagnostic insights such as tumor classification, subclassification, and grading, aiming to improve accuracy, reproducibility, and efficiency in histopathology workflows. To lower the technical barrier to researchers in accessing and utilizing advanced ML, we developed a comprehensive solution for classifying WSIs by disease or tumor subtype, based on the morphological characteristics of the tissue. Building on the NCI Cancer Research Data Commons’ (CRDC) cloud infrastructure, we take advantage of hosted data from the Imaging Data Commons (IDC) and Human Tumor Atlas Network (HTAN) paired with the computational resources available in the Cancer Genomics Cloud (CGC), powered by Velsera and funded by the NCI, to host a reproducible WSI ML solution. This solution includes data preparation and preprocessing, model training, evaluation, and predictions, within an interactive analysis environment. The analysis harnesses the computational capabilities of the CGC platform to efficiently process large-scale datasets. To facilitate data preparation, we integrated tools previously developed on the CGC to extract regions of interest (ROIs) from WSIs, enabling the creation of expanded datasets by generating tumor tissue patches for model training. To further enhance classification performance, a consensus model approach was employed—combining predictions from multiple models to improve robustness and accuracy. This work demonstrates the power of coupling AI-ready harmonized imaging data, available from within the CRDC, with advanced ML techniques, and showcases an end-to-end ML workflow on the CGC. By hosting the solution on an accessible and interactive system, users will be able to seamlessly apply the analysis to their own datasets, gaining meaningful insights with minimal technical barriers. The combination of best practices in data preparation, model development, and consensus-driven accuracy improvement makes ML for histopathology more accessible to users, enabling further research and deeper insights into cancer etiology and diagnostics. Citation Format: Jovana Babić, Nevena Nikolić, Milan Kovačević, Rowan Beck, Tariq Khoyratty, Zelia Worman. Cloud-based machine learning for enhanced tumor classification in cancer genomics: an end-to-end solution for whole slide imaging data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 7434.

The Impact of Machine Learning on Future Defense Strategies

Machine learning (ML) is increasingly pivotal in shaping future defense strategies, offering advancements across critical domains such as cybersecurity, resource allocation, and autonomous systems. This research explores the multifaceted impact of ML in defense, focusing on three primary areas: enhancing threat detection and mitigation in cybersecurity, optimizing resource allocation and logistics in military operations, and navigating the ethical and strategic implications of autonomous defense systems. The study utilizes qualitative research methods, particularly through secondary data analysis from government reports, academic publications, industry white papers, and ethical frameworks. Findings indicate that ML algorithms significantly bolster threat detection by leveraging Anomaly Detection Theory and Game Theory, thereby enhancing responsiveness to cyber threats. In military logistics, ML models informed by Operations Research and Supply Chain Management Theory optimize resource distribution, improve operational efficiencies, and support mission readiness. Ethically, the deployment of ML in autonomous defense systems prompts considerations of moral responsibilities, biases, and strategic risks, necessitating comprehensive governance frameworks. In conclusion, while ML offers transformative potential in defense strategies, effective implementation requires robust ethical guidelines, strategic foresight, and interdisciplinary collaboration to mitigate risks and maximize benefits.

Improved Alzheimer Disease Diagnosis With a Machine Learning Approach and Neuroimaging: Case Study Development.

Alzheimer disease (AD) is a severe neurological brain disorder. While not curable, earlier detection can help improve symptoms substantially. Machine learning (ML) models are popular and well suited for medical image processing tasks such as computer-aided diagnosis. These techniques can improve the process for an accurate diagnosis of AD. In this paper, a complete computer-aided diagnosis system for the diagnosis of AD has been presented. We investigate the performance of some of the most used ML techniques for AD detection and classification using neuroimages from the Open Access Series of Imaging Studies (OASIS) and Alzheimer's Disease Neuroimaging Initiative (ADNI) datasets. The system uses artificial neural networks (ANNs) and support vector machines (SVMs) as classifiers, and dimensionality reduction techniques as feature extractors. To retrieve features from the neuroimages, we used principal component analysis (PCA), linear discriminant analysis, and t-distributed stochastic neighbor embedding. These features are fed into feedforward neural networks (FFNNs) and SVM-based ML classifiers. Furthermore, we applied the vision transformer (ViT)-based ANNs in conjunction with data augmentation to distinguish patients with AD from healthy controls. Experiments were performed on magnetic resonance imaging and positron emission tomography scans. The OASIS dataset included a total of 300 patients, while the ADNI dataset included 231 patients. For OASIS, 90 (30%) patients were healthy and 210 (70%) were severely impaired by AD. Likewise for the ADNI database, a total of 149 (64.5%) patients with AD were detected and 82 (35.5%) patients were used as healthy controls. An important difference was established between healthy patients and patients with AD (P=.02). We examined the effectiveness of the three feature extractors and classifiers using 5-fold cross-validation and confusion matrix-based standard classification metrics, namely, accuracy, sensitivity, specificity, precision, F1-score, and area under the receiver operating characteristic curve (AUROC). Compared with the state-of-the-art performing methods, the success rate was satisfactory for all the created ML models, but SVM and FFNN performed best with the PCA extractor, while the ViT classifier performed best with more data. The data augmentation/ViT approach worked better overall, achieving accuracies of 93.2% (sensitivity=87.2, specificity=90.5, precision=87.6, F1-score=88.7, and AUROC=92) for OASIS and 90.4% (sensitivity=85.4, specificity=88.6, precision=86.9, F1-score=88, and AUROC=90) for ADNI. Effective ML models using neuroimaging data could help physicians working on AD diagnosis and will assist them in prescribing timely treatment to patients with AD. Good results were obtained on the OASIS and ADNI datasets with all the proposed classifiers, namely, SVM, FFNN, and ViTs. However, the results show that the ViT model is much better at predicting AD than the other models when a sufficient amount of data are available to perform the training. This highlights that the data augmentation process could impact the overall performance of the ViT model.

Abstract 7451: Efficacy of chemoradiotherapy in older patients with advanced stage head and neck cancer: A propensity and machine learning analysis

Abstract Background: The rising incidence of head and neck squamous cell carcinoma (HNSCC) is linked with both human papillomavirus and aging population. As a result, there is an emerging need for developing management strategies for older patients with advanced-stage HNSCC. This study aims to use machine learning (ML) techniques to examine the efficacy of definitive, concurrent chemoradiotherapy (CRT) for older patients (≥65 years) with stage III-IV HNSCC. We used permutation feature importance (PFI) to compare the efficacy of CRT with single-modality radiotherapy (RT) for the overall survival (OS) of this HNSCC subgroup. Finally, we compared the results obtained using ML analysis with propensity-score matched (PSM) analysis to clarify the efficacy of CRT in this patient population. Methods: A total of 3570 patients with HNSCC were obtained from the Surveillance, Epidemiology, and End Results (SEER) Program to build the ML model. We compared voting ensemble and extreme gradient boosting algorithms for stratifying older (≥65 years) HNSCC patients into OS risk groups. Furthermore, the PFI analysis was based on stacked ML models to evaluate the efficacy of CRT regarding OS of older (≥65 years) HNSCC patients. Then, we performed a 1:1 propensity-score matched (PSM) analysis to reduce treatment selection bias and analyze the prognostic role of CRT. Results: Our ML model showed a performance accuracy of 85.2% in predicting OS in this subgroup of HNSCC. The aggregate feature importance showed that CRT is the most important feature for OS in this group of HNSCC patients. In addition, patients with stage IV disease showed better survival than those with stage III. The PSM confirmed the efficacy of CRT compared to RT alone. Conclusions: This study showed that CRT may be more feasible and effective than RT alone for older patients with stage III-IV HNSCC. This further emphasized that advanced-stage treatment decisions should rather be made according to other patient-related factors than chronologic (calendar) age alone. Therefore, validating these models with multi-institutional datasets and testing them in the context of clinical trials is warranted to confirm the efficacy of CRT over RT alone for older patients (≥65 years) with stage III-IV HNSCC. Citation Format: Rasheed Omobolaji Alabi, Alhadi Almangush, Antti Mäkitie. Efficacy of chemoradiotherapy in older patients with advanced stage head and neck cancer: A propensity and machine learning analysis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 7451.

Artificial Intelligence: Applications in Pharmacovigilance Signal Management.

Pharmacovigilance is the science of collection, detection, and assessment of adverse events associated with pharmaceutical products for the ongoing monitoring and understanding of those products' safety profiles. Part of this process, signal management, encompasses the activities of signal detection, signal validation/confirmation, signal evaluation, and ultimately, final assessment as to whether a safety signal constitutes a new causal adverse drug reaction. Artificial intelligence is a group of technologies including machine learning and natural language processing that are revolutionizing multiple industries through intelligent automation. Here, we present a critical evaluation of studies leveraging artificial intelligence in signal management to characterize the benefits and limitations of the technology, the level of transparency, and our perspective on best practices for the future. To this end, PubMed and Embase were searched cumulatively for terms pertaining to signal management and artificial intelligence, machine learning, or natural language processing. Information pertaining to the artificial intelligence model used, hyperparameter settings, training/testing data, performance, feature analysis, and more was extracted from included articles. Common signal detection methods included k-means, random forest, and gradient boosting machine. Machine learning algorithms generally outperformed traditional frequentist or Bayesian measures of disproportionality per various metrics, showing the potential utility of advanced machine learning technologies in signal detection. In signal validation and evaluation, natural language processing was typically applied. Overall, methodological transparency was mixed and only some studies leveraged "gold standard" publicly available positive and negative control datasets. Overall, innovation in pharmacovigilance signal management is being driven by machine learning and natural language processing models, particularly in signal detection, in part because of high-performing bagging methods such as random forest and gradient boosting machine. These technologies may be well poised to accelerate progress in this field when used transparently and ethically. Future research is needed to assess the applicability of these techniques across various therapeutic areas and drug classes in the broader pharmaceutical industry.

Application of Machine Learning in the Pricing of Derivative Financial Instruments

The global derivatives market is experiencing significant growth, and precise pricing of derivatives is essential for optimizing their financial utilization. However, conventional derivative pricing methodologies, such as the Black-Scholes option pricing model, are predicated on strict assumptions, rendering them challenging to apply accurately in practice. In light of advancements in financial technology, the application of machine learning techniques for derivative pricing has emerged as a prominent area of scholarly inquiry. This article seeks to conduct a comprehensive literature review of machine learning-based derivative pricing methods, aiming to assess the current state of academic research in this domain and to elucidate how existing literature employs machine learning approaches for derivative pricing. The paper will initially concentrate on the methodologies associated with machine learning in derivative pricing, utilizing specific studies as illustrative examples of practical applications. Subsequently, it will compile instances where machine learning techniques have been employed for hedging and risk management in relation to derivatives. Finally, the paper will provide a synthesis of findings and offer insights into the future trajectory of machine learning applications in derivative pricing. The machine learning pricing methodology, which eschews reliance on traditional models in favor of extensive historical data analysis, holds the potential for more accurate pricing of derivative products and represents a promising direction for future development.

A method to benchmark high-dimensional process drift detection

Abstract Process curves are multivariate finite time series data coming from manufacturing processes. This paper studies machine learning that detect drifts in process curve datasets. A theoretic framework to synthetically generate process curves in a controlled way is introduced in order to benchmark machine learning algorithms for process drift detection. An evaluation score, called the temporal area under the curve, is introduced, which allows to quantify how well machine learning models unveil curves belonging to drift segments. Finally, a benchmark study comparing popular machine learning approaches on synthetic data generated with the introduced framework is presented that shows that existing algorithms often struggle with datasets containing multiple drift segments.

Abstract 2419: Impact of immune contexture on immunotherapy response in NSCLC: insights from multiplex IHC and machine learning-based phenotyping

Abstract Background: The relationship between immune phenotypes (infiltrated, excluded, and desert) and response to immunotherapy in non-small cell lung cancer (NSCLC) is an area of active investigation. Understanding the tumor immune contexture can indeed provide valuable insights into the tumor microenvironment (TME) and immune dynamics, potentially guiding personalized therapeutic strategies. Methods: We analyzed a discovery cohort of 148 NSCLC patients using the CD8/PanCK multiplex immunohistochemistry (mIHC) panel to define the immune phenotype. Tumors with &amp;lt;2% of CD8+ cells were classified as "desert." Tumors with &amp;gt;2% CD8+ cells within the tumor area were categorized as "infiltrated, " while those with &amp;gt;2% CD8+ cells restricted to the periphery were labeled "excluded." Correlations between immune phenotypes and immunotherapy responses (PFS, OS and response rates) were assessed. Validation was performed on an independent cohort of 117 patients. Results: In the discovery cohort, tumors with an "infiltrated" phenotype demonstrated significantly better responses to immunotherapy compared to "excluded" and "desert" tumors (p = 0.009 “infiltrated” vs “desert”, p=0.045 “infiltrated” vs “excluded”). Immune exclusion was associated with lack of response. These findings were confirmed in the validation cohort, reinforcing the prognostic value of immune phenotyping in NSCLC. Additionally, tertiary lymphoid structures (TLS) and macrophage infiltration, known predictors of response to immunotherapy in NSCLC, were analyzed using an ISO-compliant multiplex immunohistofluorescence (mIHF) panel. Machine learning algorithms were employed to automate immune phenotyping and produce rapid summaries of TME characteristics. The multiplex IHF panel reliably assessed immune phenotypes, TLS presence, and macrophage infiltration with high reproducibility. Automated phenotyping using machine learning enabled rapid and accurate characterization of TME features, facilitating the integration of immune phenotyping into clinical workflows. Conclusions: Tumor immune contexture significantly influences the outcomes of immunotherapy in NSCLC. "Infiltrated" tumors are associated with favorable responses, while immune exclusion predicts resistance. Our ISO-certified multiplex IHF panel, combined with machine learning automation, offers a robust and scalable approach to immune phenotyping, supporting its implementation in precision oncology. Keywords: NSCLC, immune phenotyping, multiplex IHC, immune exclusion, infiltrated tumors, tertiary lymphoid structures, macrophage infiltration, machine learning, immunotherapy response. Citation Format: Jean-Philippe Guegan, Imane Nafia, Florent Petraud, Lucile Vanhersecke, Christophe Rey, Alban Bessede, Antoine Italiano. Impact of immune contexture on immunotherapy response in NSCLC: insights from multiplex IHC and machine learning-based phenotyping [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2419.

Detection of whey protein concentrate adulteration using laser-induced breakdown spectroscopy combined with machine learning.

In recent years, food fraud issues related to whey protein supplements have disrupted the market and caused significant concern among consumers. Conventional analytical methods such as HPLC and ion exchange chromatography are commonly used to detect adulteration in whey protein supplements. However, these methods are costly, time-consuming and require specialised operation, making them less suitable for a wider range of users. This study presents a rapid and reliable approach for verifying the authenticity of whey protein supplements using laser-induced breakdown spectroscopy (LIBS) and machine learning. Specifically, this approach is employed to identify 15 brands of whey protein concentration (WPC), quantify protein and carbohydrate concentrations, distinguish three types of adulterants, and predict the level of adulteration. The relationship between LIBS data and analyte labels is established using machine learning methods, including partial least squares regression (PLSR), partial least squares discriminant analysis (PLS-DA), and kernel extreme learning machine (K-ELM). The accuracy for identifying WPC brands was over 0.977, and the highest coefficient of determination (R2) for quantifying protein and carbohydrate contents was 0.984 and 0.978, respectively. In addition, different adulterants can be differentiated with accuracies exceeding 0.986, and the R2 values for adulteration prediction are above 0.967 in most cases. These results suggest that LIBS combined with machine learning can serve as a viable and efficient solution for detecting adulteration in whey protein supplements.

Abstract 2400: Utilizing machine learning to predict clinical staging with head and neck cancer gene expression datasets

Abstract Introduction: Head and neck cancer is a deadly disease with relatively stagnant cure rates. Predictive oncology might allow treatment stratification for surgery, radiation, and chemotherapy combinations based on both stage and other risks. As more high dimensional data from genomic expression datasets become publicly available, machine learning algorithms could be used for staging/treatment refinement. In the present study, we applied machine learning algorithms to our previously published smaller U133A Affymetrix gene expression dataset and the NCI Cancer Genome Atlas (TCGA). Methods: Models were trained to classify T and N staging data for TCGA and Affymetrix chip datasets, by early (T1/2) versus advanced T stage (T3+), and for N0 versus N1+, respectively, for TCGA and Affymetrix data separately. TCGA data was split according to HPV status into two datasets. Models to predict early vs. advanced stage were trained and tested on randomly (proportional with respect to staging) split data: large dataset (HPV- T and HPV- N, n&amp;gt;=350) models were trained on 80% and tested on the unseen 20% samples, while for the remainder (n&amp;lt;=50), data were split 70%-30% for sufficient testing. Scikit-learn models utilized were: Bagged (aggregated) Support Vector Classifiers (BSVCs), Random Forest (RF), Gradient Boosting (GB), and Linear Regression (LR). LR was added unaltered as a baseline prediction, while the other three models were subjected to hyperparameter tuning (assessed for all combinations of selected settings) and feature selection: models were trained and tested on whole expression data, iteratively reduced to the top half important genes down to &amp;lt;=100 genes. Models were assessed by test accuracy and area under receiver operator characteristic curve (AUC-ROC), then hyperparameter-tuned models and respective gene lists were assessed over 50 trials (random initializations and data splits). Results: Hyperparameter-tuned RF and BSVCs models with selected gene lists significantly outperformed LR in 6/6 tasks, while GB outperformed LR in 2/6 tasks. Example highest performing models (test accuracy [95% CI], AUC-ROC [95% CI]) were: For Affymetrix N, BSVCs (0.893 [0.872-0.915], 0.95 [0.931-0.969] outperformed LR (0.49 [0.453-0.527], 0.476 [0.435-0.517]). For TCGA HPV- N, BSVCs (0.832 [0.818-0.847], 0.892 [0.877-0.907]) outperformed LR (0.581 [0.568-0.595], 0.615 [0.6-0.631]). For TCGA HPV+ T, RF (0.835 [0.815-0.854], 0.954 [0.942-0.965]) outperformed LR (0.669 [0.645-0.694], 0.76 [0.732-0.789]). Conclusions: Utilizing this hyperparameter tuning and feature selection approach, we have identified machine learning techniques that can be used to analyze high dimensional data of small and large genomic expression datasets to predict clinical features and identify predictive genes for further analysis, applicable to head and neck cancer, a relatively rare malignancy (about 3% of total cancers). Citation Format: Nathan DeMichaelis, Amrit Menon, Dalton Schutte, Frank G. Ondrey. Utilizing machine learning to predict clinical staging with head and neck cancer gene expression datasets [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2400.

Abstract 7438: Mapping T cell reactivity against the human genome using machine learning uncovers target specific TCRs against novel tumour antigens

Abstract Introduction: T cell engagers (TCEs) are a class of targeted therapies with demonstrated clinical potential in solid tumors. A main challenge currently facing TCEs is the availability of clean cancer targets with a wide therapeutic window. Intracellular targets presented to T cells by HLA complexes are strongly linked to disease mechanisms and are therefore very attractive, but drugging HLA-presented targets has been challenged by our ability to map the specificity of T cell repertoires. We here present the T-Cα platform, which represents a fundamental change in TCR discovery methodology, enabling us to discover &amp;gt;10, 000 TCRs per week against 100s of targets. This data has enabled us to train machine learning (ML) models and to demonstrate the ability of these models to extrapolate beyond the training data. In effect, this results in the discovery of novel therapeutically relevant TCRs and enhances our understanding of the interface between T cell reactivity and the human genome. Experimental Procedures: We used the T-Cα platform to discover more than 50 novel targets across all areas of the genome, including regions previously considered non-coding. We screened billions of T cells from 100s of donors against these targets and other targets of interest. We generated single-cell transcriptomic data as part of the screens and predicted the potency of each TCR at the discovery stage. For the purposes of model development, we generated matched negative data with non-binding TCRs during the course of the screens. These datasets, representing 10, 000s of TCR-epitope pairs, were collated alongside available public data and a machine learning model was trained to predict TCR specificity. The TCRs were functionally validated downstream for potency and cross-reactivity. Results: We have generated, to our knowledge, the largest high-quality database of TCR-epitope pairs against 100s of targets. We demonstrate that with this data, alongside matched negatives, we can train models that are able to predict TCR specificity with high concordance with known epitope pairings. We used the model to predict the specificity of unseen TCRs against novel targets of therapeutic interest in oncology. We further demonstrate that the model can extrapolate beyond the training data and identify novel target-specific TCRs suitable for the development of TCEs. Conclusions: We have developed a uniquely high-throughput, high-content TCR discovery platform with which we are able to discover &amp;gt;10, 000 specific TCRs per week against 100s of targets. We have trained a machine learning model which can accurately extrapolate to unseen TCRs, enhancing our understanding of T cell reactivity against the human genome. With the T-Cα platform, we have built a pipeline of novel T cell engagers for the treatment of solid tumors, the first of which we expect to enter IND enabling studies mid 2025. Citation Format: Nathaniel J. Davies, Amalia Martinez, Maria Busz, David Cook, Xiaoyan Pan, Juan Bolivar, Simon Wright, Eleanor Bagg, Graham Ogg, Sophie Johnson, Sophie Wells, Sophie Richard, Samantha Drennan, Samuel Ward, Sarah Leonard, Thomas L. Andresen. Mapping T cell reactivity against the human genome using machine learning uncovers target specific TCRs against novel tumour antigens [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 7438.

Revolutionizing Fraud Detection in Finance through Machine Learning

In this paper, the focus will be on the employment of machine learning technology in the identification and combating of financial transaction fraud. It can be seen that with the development of the financial industry’s digital environment, the tools necessary for financial transaction fraud are becoming multifaceted, threatening individuals, enterprises, and financial entities. The traditional approach of fraud detection is now proving unsuitable for dealing with new forms of fraud because of their very nature. As we know, machine learning has powerful data processing ability, complicated pattern recognition ability, self-learning and adaptation, and so on, so people look forward to adopting machine learning to fight financial transaction fraud. The case studies are given to explain how the various ML techniques can enhance the efficiency and accuracy of fraud detection in credit card fraud, account hijacking, and money laundering. However, there is the problem of the quality of the data used, protection of user data, ability to explain the results of the machine learning model, costs involved in implementing the system, and the compatibility of the system with the current systems. It is expected that with the growth of machine learning and the technologies associated with it, it will play an even greater role in the field of financial security and thus shape a more safe, efficient, and intelligent financial environment.