Challenging behaviors in children with autism is a serious clinical condition, oftentimes leading to aggression or self-injurious actions. The Revised Family Observation Schedule 3 rd Edition (FOS-R-III) is an intensive and fine-grained scale used to observe and analyze the behaviors of individuals with autism, which facilitates the diagnosis and monitoring of autism severity. Previous AI-based approaches for automated behavior analysis in autism often focused on predicting facial expressions and body movements without generating a clinically meaningful scale, mostly utilizing visual information. In this study, we propose a deep-learning based algorithm with audio-visual multimodal-data clinically coded with the FOS-R-III, named AV-FOS model. Our proposed AV-FOS model leverages transformer-based structure and self-supervised learning to intelligently recognize Interaction Styles (IS) in the FOS-R-III scale from subjects' video recordings. This enables the automatic generation of the FOS-R-III measures with clinically acceptable accuracy. We explore the IS recognition using a multimodal large language model, GPT4V, with prompt engineering provided with FOS-R-III measure definitions as the baseline for this study and compare with other vision-based deep learning algorithms. We believe this research represents a significant advancement in autism research and clinical accessibility. The proposed AV-FOS and our FOS-R-III dataset will serve as a gateway toward the digital health era for future AI models related to autism.
Background: AI-enabled digital twins (DTs) are advanced virtual models of a complex real-world system, which have the potential to transform clinical decision-making. Despite the growing interest in such DTs, the literature lacks a unified framework for their development and implementation. Objective: This study aims to map the existing knowledge on AI-enabled DTs for clinical decision-making, and develop a comprehensive framework for their development and implementation. Methods: Informed by frameworks established by Arksey and O'Malley, and the Joanna Briggs Institute, we performed a scoping review of studies on the development and implementation of AI-enabled DTs for clinical decision-making in any healthcare setting. The search strategy was developed by a librarian for three databases from the date of inception until August 2023. We also conducted a grey literature search on Google Scholar. One reviewer screened titles and abstracts, full-text articles, and charted data, and the second reviewer verified them. Quantitative data were summarized using frequency and proportions, and qualitative data were summarized using content analysis. Key steps in DT development were identified to create the DECIDE-Twin framework. Results: Eleven articles were included: seven reviews and four empirical studies. The reviews contained either a framework or information that was used to construct our comprehensive framework. The empirical studies reported the DT development, and one reported a common infrastructure for a wide range of DT applications. Conclusion: We developed the DECIDE-Twin framework that could serve as a guide for researchers and practitioners in DT development and implementation for clinical decision-making. Further research is needed to validate and implement this framework for various clinical applications.
Pediatric pneumonia is the leading cause of death among children under five years worldwide, imposing a substantial burden on affected families. Currently, there are three significant hurdles in diagnosing and treating pediatric pneumonia. Firstly, pediatric pneumonia shares similar symptoms with other respiratory diseases, making rapid and accurate differential diagnosis challenging. Secondly, primary hospitals often lack sufficient medical resources and experienced doctors. Lastly, providing personalized diagnostic reports and treatment recommendations is labor-intensive and time-consuming. To tackle these challenges, we proposed a Medical Multimodal Large Language Model for Pediatric Pneumonia (P2Med-MLLM). It was capable of handling diverse clinical tasks—such as generating free-text medical records and radiology reports—within a unified framework. Specifically, P2Med-MLLM was trained on a large-scale dataset, including real clinical information from 163,999 outpatient and 8,684 inpatient cases. It can process both plain text data (e.g., outpatient and inpatient records) and interleaved image-text pairs (e.g., 2D chest X-ray images, 3D chest Computed Tomography images, and corresponding radiology reports). We designed a three-stage training strategy to enable P2Med-MLLM to comprehend medical knowledge and follow instructions for various clinical decision-support tasks. To rigorously evaluate P2Med-MLLM's performance, we conducted automatic scoring by the large language model and manual scoring by the specialist on the test set of 642 samples, meticulously verified by pediatric pulmonology specialists. The results demonstrated the reliability of automated scoring and the superiority of P2Med-MLLM. This work plays a crucial role in assisting doctors with prompt diagnosis and treatment planning, reducing severe symptom mortality rates, and optimizing the allocation of medical resources.
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.
Front Biosci (Landmark Ed). 2025 Aug 30;30(8):44091. doi: 10.31083/FBL44091.
ABSTRACT
Zhang et al.'s recent article utilizes comprehensive single-cell data to identify differences in tumor cell populations, highlighting the CKS1B+ malignant cell subcluster as a potential target for immunotherapy. It develops a prognostic and immunotherapeutic signature (PIS) based on this subcluster, demonstrating good performance in predicting lung adenocarcinoma (LUAD) prognosis. The study also validates the role of PSMB7 in LUAD progression. However, there are areas for improvement. There is a lack of clarity regarding the relationship between the CKS1B+ malignant cell subcluster and the PIS, particularly in terms of why PSMB7 was selected for functional studies. The sequencing data are retrospectively obtained from public databases and lack prospective clinical validation. It is suggested to collect LUAD patient tissues for RT-qPCR and RNA-seq analysis and seek external multi-center validations. Additionally, integrating emerging multi-omics methods is recommended to further validate the findings. Despite these limitations, the study represents progress in understanding LUAD and treatment strategies, and continuous evaluation and refinement of multi-omics and machine learning methods are expected for future research and clinical practice.
Hugging Face has released AI Sheets, an open-source application designed to let users build, transform, and enrich datasets using AI models through a spreadsheet-like interface. The tool, available both on the Hub and for local deployment, allows users to experiment with thousands of open models, including OpenAI’s gpt-oss, without requiring code.
Artificial intelligence is consuming enormous amounts of energy, but researchers at the University of Florida have built a chip that could change everything by using light instead of electricity for a core AI function. By etching microscopic lenses directly onto silicon, they’ve enabled laser-powered computations that cut power use dramatically while maintaining near-perfect accuracy.
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.
DOVER, N.H. — Not long after he woke from surgery in June, Bill Stewart made a pact with his newest organ. He wasn’t sure how long the thing would last. The doctors had been up-front from the get-go: It could be three months or six, one year or four. Still, the uncertainty hit him as he started getting back preliminary lab results, which were okay but left room for improvement. “My pig kidney and I had a little conversation while I was laying there. I just basically said, ‘I’m going to do everything I can to make sure that you stay healthy, and I appreciate you doing everything you can to keep me upright and breathing,” Stewart said.
It’s been almost three months, and Stewart is home, back to work, and has even been able to go e-biking on a lakeside trail with his wife, blessedly untethered to the grueling schedules of dialysis for the first time in years, all thanks to a gene-edited Yucatan miniature pig named Lavender.
Stewart is the most recent recipient of a pig kidney — but chances are, he won’t hold that distinction for long. On Monday, eGenesis, a Cambridge-based biotechnology company, announced that it had been cleared by the Food and Drug Administration to begin a trial of kidneys from donor pigs that have been CRISPR’d to make their organs more human-friendly. Now, Massachusetts researchers will be performing more surgeries like Stewart’s to see whether these animal parts could serve as a lifeline for people with end-stage renal disease.
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
Introducing: our 35 Innovators Under 35 list for 2025
The world is full of extraordinary young people brimming with ideas for how to crack tough problems. Every year, we recognize 35 such individuals from around the world—all of whom are under the age of 35.
These scientists, inventors, and entrepreneurs are working to help mitigate climate change, accelerate scientific progress, and alleviate human suffering from disease. Some are launching companies while others are hard at work in academic labs. They were selected from hundreds of nominees by expert judges and our newsroom staff.
Get to know them all—including our 2025 Innovator of the Year—in these profiles.
Why basic science deserves our boldest investment
—Julia R. Greer is a materials scientist at the California Institute of Technology, a judge for MIT Technology Review’s Innovators Under 35 and a former honoree (in 2008).
A modern chip the size of a human fingernail contains tens of billions of silicon transistors, each measured in nanometers—smaller than many viruses. These tiny switches form the infrastructure behind nearly every digital device in use today.
Much of the fundamental understanding that moved transistor technology forward came from federally funded university research. But that funding is under increasing pressure, thanks to deep budget cuts proposed by the White House.
These losses have forced some universities to freeze graduate student admissions, cancel internships, and scale back summer research opportunities—making it harder for young people to pursue scientific and engineering careers.
In an age dominated by short-term metrics and rapid returns, it can be difficult to justify research whose applications may not materialize for decades. But those are precisely the kinds of efforts we must support if we want to secure our technological future. Read the full story.
The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 The US is considering annual chip supply permits in China For South Korean companies Samsung and SK Hynix, specifically. (Bloomberg $) + US lawmakers still hold power over chips in China. (CNN)
2 America has recorded its first case of screwworm in over 50 years And the warming climate is making it easier for the flies to thrive. (Vox) + Experts fear an approaching public health emergency. (The Guardian)
3 Drone warfare is dominating Ukraine’s frontline Amid relentless assaults, overhead and land drones are being put to work. (The Guardian) + How cutting-edge drones forced land-locked tanks to evolve. (NYT $) + On the ground in Ukraine’s largest Starlink repair shop. (MIT Technology Review)
4 OpenAI is working out why chatbots hallucinate so much Examining a model’s incentives provides some clues. (Insider $) + Models’ tendency to confidently present falsehoods as fact is a big problem. (TechCrunch) + Why does AI hallucinate? (MIT Technology Review)
5 How one man is connecting Silicon Valley to the Middle East’s AI boom If you want to build a data center, Zachary Cefaratti is your man. (FT $) + The data center boom in the desert. (MIT Technology Review)
6 The first OpenAI-backed movie is coming to theaters next year The animated Critterz is hoping for a Cannes Film Festival debut. (WSJ $) + A Disney director tried—and failed—to use an AI Hans Zimmer to create a soundtrack. (MIT Technology Review)
7 Who wants to live forever? These billionaires are confident their cash will pave the way to longer lives. (WSJ $) + Putin says organ transplants could grant immortality. Not quite. (MIT Technology Review)
8 Tesla isn’t focused on selling cars any more The company’s latest Master Plan is all about humanoid robots. (The Atlantic $) + The board is willing to offer Musk a $1 trillion pay package if he delivers. (Wired $) + Uber is gearing up to test driverless cars in Germany. (The Verge) + China’s EV giants are betting big on humanoid robots. (MIT Technology Review)
9 Do aliens go on holiday? Scientists wonder whether tourism could be a potential drive for them to visit us. (New Yorker $) + How these two UFO hunters became go-to experts on America’s “mystery drone” invasion. (MIT Technology Review)
10 Vodafone’s new TikTok influencer isn’t real It’s yet another example of AI avatars being used in ads. (The Verge) + Synthesia’s AI clones are more expressive than ever. Soon they’ll be able to talk back. (MIT Technology Review)
Quote of the day
“Silicon Valley totally effed up in overhyping LLMs.”
—Palantir CEO Alex Karp criticizes those who fueled the AI hype around large language models, Semafor reports.
One more thing
Puerto Rico’s power struggles
On the southeastern coast of Puerto Rico lies the country’s only coal-fired power station, flanked by a mountain of toxic ash. The plant, owned by the utility giant AES, has long plagued this part of Puerto Rico with air and water pollution.
Before the coal plant opened Guayama had on average just over 103 cancer cases per year. In 2003, the year after the plant opened, the number of cancer cases in the municipality surged by 50%, to 167.
In 2022, the most recent year with available data, cases hit a new high of 209. The question is: How did it get this bad? Read the full story.
—Alexander C. Kaufman
We can still have nice things
A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)
+ What’s up with tennis players’ strange serving rituals? + If constant scrolling is turning your hands into gnarled claws, this stretch should help. + How to land a genuine bargain on Facebook Marketplace. + This photographer tracks down people who featured in pictures decades before, and persuades them to recreate their poses. Heartwarming stuff
Antibody–drug conjugates (ADCs) are promising targeted cancer therapies but have a limited payload scope. Antibody–bottlebrush prodrug conjugates offer modular synthesis and high drug-to-antibody ratios, enabling the use of a broad range of payloads, including lower potency drugs, while performing favorably compared to traditional ADCs in preclinical models.
Liquid biopsies, particularly those involving circulating tumor DNA (ctDNA) from patient blood, have emerged as crucial and minimally invasive adjuncts to standard tissue-based testing. ctDNA testing enables the identification of actionable mutations for targeted therapy and can be routinely used when tissue samples are unavailable for genotyping. Compared to tissue-based testing, ctDNA testing has the advantages of capturing spatial or temporal genomic heterogeneity and facilitating repeated assessments. The utility of liquid biopsies extends to multiple clinical applications, including cancer diagnosis, treatment monitoring, and minimal residual disease (MRD) detection. Numerous clinical trials are currently evaluating treatment strategies using ctDNA testing. In particular, the implementation of adjuvant treatment escalation or de-escalation based on MRD detection could dramatically transform future approaches to solid tumor treatment. Various ctDNA assays have been developed, and it is important to understand their strengths and weaknesses for effective clinical applications. Furthermore, ctDNA testing faces several technical challenges, including low sensitivity in detecting copy number alterations and fusions, as well as the possibility of detecting mutations associated with clonal hematopoiesis of indeterminate potential. In this review, we comprehensively discuss the methodologies and recent advancements in cfDNA-based liquid biopsies for cancer patients, covering diagnosis, genomic profiling, and treatment monitoring. Furthermore, we explore clinical trial designs employing ctDNA testing and anticipate forthcoming changes in patient care.
Int J Biol Macromol. 2025 Sep 5;327(Pt 2):147473. doi: 10.1016/j.ijbiomac.2025.147473. Online ahead of print.
ABSTRACT
Osteosarcoma (OS), the most prevalent primary bone malignancy in adolescents, is characterized by aggressive progression and early metastasis. However, the epigenetic drivers of its metastatic heterogeneity remain poorly understood. Herein, we integrated bulk DNA methylation profiling and single-cell RNA sequencing (scRNA-seq) to elucidate the epigenetic mechanisms driving OS metastatic heterogeneity. Consensus clustering identified two methylation subtypes (K = 2) with distinct survival outcomes, where hypermethylated (MSO-high) tumors exhibited poor prognosis. Weighted gene co-expression network analysis (WGCNA) revealed methylation-associated modules enriched in metabolic and immune pathways, pinpointing key genes such as CAMK1G and SLC11A1. Single-cell profiling uncovered MSO-high myeloid cells associated with inflammatory and oxidative phosphorylation pathways, while MSO-high OS cells displayed transdifferentiation toward fibroblasts via pseudotime trajectories, remodeling the extracellular matrix (ECM) to facilitate lung metastasis. Conversely, MSO-low tumors activated HLA-B-mediated neutrophil-CD8+ T cell interactions, promoting lymphatic metastasis via CXCR4/CXCL12 signaling. Furthermore, functional validation using the DNA demethylating agent decitabine demonstrated reduced fibroblastic transdifferentiation and suppressed invasive capacity in MSO-high osteosarcoma cells, supporting the therapeutic potential of targeting methylation dysregulation. These findings establish a model where DNA methylation dictates metastatic phenotypes through differential tumor-stromal crosstalk, providing novel targets for epigenetic therapy to disrupt fibrotic-immune networks and metastatic colonization.
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.
Login
