Liquid biopsy, specifically circulating tumor DNA (ctDNA) analysis, has emerged as a transformative tool in precision oncology, providing real-time, minimally invasive characterizations of the tumor and tumor dynamics. While tissue biopsy is a critical tool for baseline diagnosis of malignancy, it is often limited by sampling constraints and an inability to capture tumor heterogeneity. In this study, we explored the clinical utility of serial ctDNA testing in guiding therapeutic decisions across a cohort of 30 patients with diverse solid tumors. Our real-world analysis demonstrates that ctDNA profiling meaningfully influenced treatment escalation, de-escalation, disease monitoring, and early relapse prediction. Cases where ctDNA positivity indicated minimal residual disease prompted timely escalation of therapy, while ctDNA clearance allowed safe treatment de-intensification, minimizing toxicity without compromising outcomes. Longitudinal ctDNA monitoring provided a dynamic, non-invasive method for assessing treatment response and detecting recurrence months before radiological progression. Our study highlights the potential of integrating liquid biopsy into routine clinical practice to enable dynamic treatment monitoring, early detection of therapeutic resistance, and more informed, personalized decision-making across various cancer types.
Background: Adverse drug reactions (ADRs) pose significant challenges in healthcare, where early prevention is vital for effective treatment and patient safety. Objective: Traditional supervised learning methods are limited in addressing healthcare data, which is often unstructured, heavily regulated, and involves restricted access to sensitive personal information. Methods: The integration of Federated Learning (FL) and Large Language Model (LLM) offers a promising solution to these challenges since FL supports the distributed training on edge device with limited resources and the capability of LLM to deal with unstructured healthcare data. Additionally, client models trained on the edge device can be merged into a global model on the server, preserving data privacy. Results: Natural Language Processing (NLP) technologies underpinning LLM provide a full set of tools that can readily be used to process unstructured ADR as input, enabling LLM to predict ADR outcome effectively. The ADR output space can be discrete labels, unstructured texts, or both. Conclusions: This review presents a scoping review following the PRISMA protocol on the applications of Federated Large Language Model (FedLLM) in ADR prediction, aiming to explore future research venue on ADR applications
This Review of the WHO’s International Clinical Trials Registry Platform presents a snapshot of the global cancer trial landscape and provides critical empirical evidence to inform policy, practice and investment.
Liam — not his real name — is a 75-year-old retired teacher in Boston. Two years ago, his son-in-law gave him an Apple Watch. Soon after, it began flagging something strange: possible atrial fibrillation.
bioRxiv [Preprint]. 2025 Aug 25:2025.08.21.671519. doi: 10.1101/2025.08.21.671519.
ABSTRACT
Lung cancer is the most common cause of cancer-related death worldwide. Recent advancements in targeted therapies and immunotherapies have achieved remarkable success. However, patient responses to treatments with lung cancer vary substantially. The mutation status of driver genes can direct personalized treatment, but their prognostic value and treatment efficacy are limited. In this study, we developed a statistical framework named Genomic Aberration-Derived Signature for Patient Stratification (GASPS) to characterize the transcriptomic deregulation of driver genomic aberrations and stratify patients. By applying GASPS to The Cancer Genome Atlas Lung Adenocarcinoma (TCGA-LUAD) data, we developed gene signatures for 38 driver genomic aberrations, including gene mutations, amplifications, and deletions. These signatures were applied to independent lung cancer transcriptomic datasets containing a total of 2,226 patient samples. Our results indicated that these driver gene signatures are much more prognostic than their corresponding genomic mutations. Interestingly, the two EGFR-related signatures characterizing EGFR mutation and amplification, respectively, exhibited contrasting associations with prognosis, treatment response, and immune infiltration in the tumor microenvironment. Moreover, the STK11 mutation signature, rather than the mutation status, was found to be predictive of the response and long-term benefit of patients treated with immune checkpoint blockade therapy in lung cancer. This framework is readily applicable to most cancer types using existing data to improve prognostic risk assessment and treatment efficacy by guiding personalized therapies.
bioRxiv [Preprint]. 2025 Aug 25:2025.08.21.671519. doi: 10.1101/2025.08.21.671519.
ABSTRACT
Lung cancer is the most common cause of cancer-related death worldwide. Recent advancements in targeted therapies and immunotherapies have achieved remarkable success. However, patient responses to treatments with lung cancer vary substantially. The mutation status of driver genes can direct personalized treatment, but their prognostic value and treatment efficacy are limited. In this study, we developed a statistical framework named Genomic Aberration-Derived Signature for Patient Stratification (GASPS) to characterize the transcriptomic deregulation of driver genomic aberrations and stratify patients. By applying GASPS to The Cancer Genome Atlas Lung Adenocarcinoma (TCGA-LUAD) data, we developed gene signatures for 38 driver genomic aberrations, including gene mutations, amplifications, and deletions. These signatures were applied to independent lung cancer transcriptomic datasets containing a total of 2,226 patient samples. Our results indicated that these driver gene signatures are much more prognostic than their corresponding genomic mutations. Interestingly, the two EGFR-related signatures characterizing EGFR mutation and amplification, respectively, exhibited contrasting associations with prognosis, treatment response, and immune infiltration in the tumor microenvironment. Moreover, the STK11 mutation signature, rather than the mutation status, was found to be predictive of the response and long-term benefit of patients treated with immune checkpoint blockade therapy in lung cancer. This framework is readily applicable to most cancer types using existing data to improve prognostic risk assessment and treatment efficacy by guiding personalized therapies.
Background: The effective implementation of personalized pharmacogenomics (PGx) requires the integration of released clinical guidelines into decision support systems (CDSS) to facilitate clinical applications. Large language models (LLMs) can be valuable tools for automating information extraction and updates. Objective: To assess the effectiveness of repeated cross-comparisons and an agreement-threshold strategy in two advanced LLMs as supportive tools for updating information. Methods: The study evaluated the performance of two LLMs, GPT-4o and Gemini-1.5-Pro, in extracting PGx clinical guidelines and comparing their outputs with expert-annotated evaluations. The two LLMs classified 385 PGx clinical guidelines, with each recommendation tested 20 times per model. Accuracy was assessed by comparing the results with manually labeled data. Two prospectively defined strategies were employed to identify inconsistent predictions. The first involved repeated cross-comparison, flagging discrepancies between the most frequent classifications from each model. The second employed a consistency threshold strategy, which designated predictions appearing in less than 60% of the 40 combined outputs as unstable. Cases flagged by either strategy were subjected to manual review. This study also estimated the overall cost of model usage and was conducted between October 1 and November 30, 2024. Results: GPT-4o and Gemini-1.5-Pro yielded reproducibility rates of 97.8% (7,534/7,700) and 98.9% (7,612/7,700), respectively, based on the most frequent classification for each query. Compared with expert labels, GPT-4o achieved 93.5% accuracy (Cohen’s Kappa=0.90; P<.001 and gemini-1.5-pro accuracy kappa="0.89;" p both models demonstrated high overall performance with comparable weighted average f1 scores gemini: the generated consistent predictions for of guideline items reducing need manual review by among these agreed-upon cases only one diverged from expert labels. applying a predefined agreement-threshold strategy further reduced number priority to although error rate slightly increased inconsistencies identified through methods prompted prioritization minimize errors enhance clinical applicability. total combined cost using llms was conclusions: findings suggest that two can effectively streamline pgx integration into cdss while maintaining minimal cost. selective remains necessary this approach offers practical scalable solution classification in workflows.>
In lieu of traditional genetic variant testing approaches, an approach using scalable variant classification in primary human T cells with a clinically relevant readout can inform rapid diagnosis and treatment of inborn errors of immunity.
Single-cell transcriptomics analysis and multimodal profiling (STAMP) by imaging enables single-cell analysis of cells in suspension without the need for sequencing. The markedly reduced costs and flexible experimental designs support the profiling of millions of cells or the large-scale multiplexing of conditions, perturbations, and sample types.
DNA-methylation and gene-expression profiling of tissue sections at near single-cell resolution can be used to create detailed spatial maps showing how methylation and transcription interact to shape cell identity and tissue development.
Artificial intelligence is fundamentally reshaping how the world operates. With its potential to automate repetitive tasks, analyze vast datasets, and augment human capabilities, the use of AI technologies is already driving changes across industries.
In health care and pharmaceuticals, machine learning and AI-powered tools are advancing disease diagnosis, reducing drug discovery timelines by as much as 50%, and heralding a new era of personalized medicine. In supply chain and logistics, AI models can help prevent or mitigate disruptions, allowing businesses to make informed decisions and enhance resilience amid geopolitical uncertainty. Across sectors, AI in research and development cycles may reduce time-to-market by 50% and lower costs in industries like automotive and aerospace by as much as 30%.
“This is one of those inflection points where I don’t think anybody really has a full view of the significance of the change this is going to have on not just companies but society as a whole,” says Patrick Milligan, chief information security officer at Ford, which is making AI an important part of its transformation efforts and expanding its use across company operations.
Given its game-changing potential—and the breakneck speed with which it is evolving—it is perhaps not surprising that companies are feeling the pressure to deploy AI as soon as possible: 98% say they feel an increased sense of urgency in the last year. And 85% believe they have less than 18 months to deploy an AI strategy or they will see negative business effects.
Companies that take a “wait and see” approach will fall behind, says Jeetu Patel, president and chief product officer at Cisco. “If you wait for too long, you risk becoming irrelevant,” he says. “I don’t worry about AI taking my job, but I definitely worry about another person that uses AI better than me or another company that uses AI better taking my job or making my company irrelevant.”
But despite the urgency, just 13% of companies globally say they are ready to leverage AI to its full potential. IT infrastructure is an increasing challenge as workloads grow ever larger. Two-thirds (68%) of organizations say their infrastructure is moderately ready at best to adopt and scale AI technologies.
Essential capabilities include adequate compute power to process complex AI models, optimized network performance across the organization and in data centers, and enhanced cybersecurity capabilities to detect and prevent sophisticated attacks. This must be combined with observability, which ensures the reliable and optimized performance of infrastructure, models, and the overall AI system by providing continuous monitoring and analysis of their behavior. Good quality, well-managed enterprise-wide data is also essential—after all, AI is only as good as the data it draws on. All of this must be supported by AI-focused company culture and talent development.
This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. It was researched, designed, and written entirely by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.
PURPOSE OF REVIEW: Diagnosing sarcoidosis remains challenging. Histology findings and a variable clinical presentation can mimic other infectious, malignant, and autoimmune diseases. This review synthesizes current evidence on histopathology, sampling techniques, imaging modalities, and biomarkers and explores how emerging 'omics' and artificial intelligence tools may sharpen diagnostic accuracy.
RECENT FINDINGS: Within the typical granulomatous lesions, limited or 'burned-out' necrosis is an ancillary finding, which can be present in up to one-third of sarcoid biopsies, and demands a careful differential diagnostic work-up. Endobronchial ultrasound-guided transbronchial needle aspiration of lymph nodes has replaced mediastinoscopy as first-line sampling tool, while cryobiopsy is still under validation. Volumetric PET metrics such as total lung glycolysis and somatostatin-receptor tracers refine activity assessment; combined FDG PET/MRI improves detection of occult cardiac disease. Advanced bronchoalveolar lavage (BAL) immunophenotyping via flow cytometry and serum, BAL, and genetic biomarkers show to correlate with inflammatory burden but have low diagnostic value. Multi-omics signatures and Positron Emission Tomography with Computer Tomography radiomics, supported by deep-learning algorithms, show promising results for noninvasive diagnostic confirmation, phenotyping, and disease monitoring.
SUMMARY: No single test is conclusive for diagnosing sarcoidosis. An integrated, multidisciplinary strategy is needed. Large, multicenter, and multiethnic studies are essential to translate and validate data from emerging AI tools and -omics research into clinical routine.
Background: Health care continues to advance through digital innovation, and technology-enabled processes and interventions are increasingly being introduced to deliver and expand access to care. In this evolving digital health ecosystem, health care professionals (HCPs), learners, and organizations may not be prepared or equipped with the knowledge, skills, and behaviors required to navigate these new digital tools while simultaneously sustaining and integrating compassionate care. Moreover, the tools may not be designed and implemented in a manner that facilitates digital compassion. Objective: This study aimed to identify (1) core digital compassion competencies for health professionals and (2) digital compassion health IT attributes. Methods: We conducted this study based on the Delphi method, a consensus-building technique using structured group communication that allows a group of experts to identify competencies and agree on items such as standards and attributes by achieving consensus on a given topic. To encourage enriched discussions, we used a modified eDelphi method, where the first round consisted of a group activity and focus group rather than a questionnaire. Due to COVID-19 pandemic restrictions, the first round was held online. Subsequent rounds consisted of questionnaires administered via email and a web-based survey. Using purposive sampling, participants were recruited from project partners and networks of the research team. A panel of experts across Canada in the fields of compassion, health professional or medical education, and technology was engaged to identify and prioritize professional domains and competency statements, as well as essential attributes for the development and deployment of digital technologies for compassionate care. Results: A total of 54 experts across Canada were recruited, representing diverse professions including patients or service users, HCPs, administrators, policy makers, health educators, data scientists, health technology designers, and software engineers. Overall, 9 focus groups were conducted and analyzed thematically. Seven domains of digital compassion were identified: (1) digital literacy, (2) patient preference, (3) collaboration and co-design, (4) therapeutic relationship, (5) ethical implications, (6) patient safety, and (7) technology safety. Technology attributes to facilitate digital compassion were also generated. We reached consensus after several subsequent rounds, resulting in 58 digital compassion competency statements and 15 technology attributes. Conclusions: This study identified a digital compassion framework consisting of competencies for HCPs and attributes for digital technologies that would enhance compassion in virtual care encounters. To promote a cultural shift where technologies are perceived to be not only efficient but also compassionate, practices of co-design, training, and ongoing evaluation and iteration must be prioritized within health care organizations. Future research should explore the adaptability of the professional competencies and technology attributes to specific medical specialties or in patient populations.
Immune checkpoint blockade (ICB) has improved outcomes for patients with head and neck squamous cell carcinoma (HNSCC), but predictive biomarkers remain limited. Here, we use a time-resolved, multi-omic approach in a murine HNSCC model to characterize peripheral immune responses to ICB. Single-cell transcriptomics and T/B cell receptor analyses reveal early on-treatment expansion of effector memory T and B cell repertoires in responders, preceding tumor regression. These dynamic immune features inform a composite transcriptional signature that accurately predicts ICB response in independent human HNSCC cohorts. LiBIO outperforms existing biomarkers and generalizes to melanoma, non-small cell lung cancer, and breast cancer without retraining. These findings suggest that early treatment-induced changes in circulating immune repertoires reflect the host's capacity to mount an effective antitumor response. This work provides a framework for leveraging transient peripheral immune dynamics to develop non-invasive, high-fidelity biomarkers for response to immunotherapy across cancer types.
Background: Chatbots have demonstrated promising capabilities in Medicine, scoring passing grades for board examinations across various specialties. However, their tendency to express high levels of confidence in their responses, even when incorrect, poses a limitation to their utility in clinical settings. Objective: To examine whether token probabilities outperform chatbots' Expressed Confidence levels in predict-ing the accuracy of their responses to medical questions. Methods: Seven large languages models (LLMs), comprising both commercial (GPT-3.5, GPT-4 and GPT-4o) and open-source (Llama 3-8b, Llama 3-70b, Phi-3-Mini, and Phi-3-Medium), were prompted to respond to a set of 2,522 questions from the US Medical Licensing Examination (MedQA database). Addition-ally, the models rated their confidence from 0 to 100 and the token probability of each response was extracted. The models’ success rates were measured, and the predictive performances of both Ex-pressed Confidence and Response Token Probability in predicting response accuracy were evaluated using Area Under the Receiver Operating Characteristic Curve (AUROC), Adapted Calibration Error (ACE) and Brier score. Sensitivity analyses were conducted using additional questions sourced from other databases in English (MedMCQA, n=2,797), Chinese (MedQA Main-land China, n=3,413 and Taiwan, n=2,808), and French (FrMedMCQA, n=1,079). Results: Overall, mean accuracy ranged from 52.7%[50.8-54.7] for Phi-3-Mini to 87.6%[86.2-88.9] for GPT-4o. Across the US Medical Licensing Examination questions, all chatbots consistently expressed high levels of confidence in their responses (ranging from 90[90-90] for Llama 3-70B to 100[100–100] for GPT-3.5). However, Expressed Confidence failed to predict response accuracy (AUROC ranging from 0.52[0.50-0.53] for Phi 3 Mini to 0.68[0.65-0.71] for GPT-4o). In contrast, the Response Token Probability consistently outperformed Expressed Confidence for predicting response accuracy (AU-ROC ranging from 0.67[0.65-0.69] for Phi-3-Mini to 0.83[0.81-0.85] for Llama 3-70B, all p-values
Login
