A personalized DNA vaccine targeting up to 40 patient-specific neoantigens generated robust immune responses and encouraging survival outcomes in patients with MGMT-unmethylated glioblastoma in a small Phase I clinical trial, according to new findings published in Nature Cancer.
The study evaluated GNOS-PV01, a personalized therapeutic cancer vaccine developed by Geneos Therapeutics in collaboration with researchers at Washington University School of Medicine in St. Louis. Investigators reported that the vaccine was safe, feasible to administer, and capable of stimulating circulating and tumor-infiltrating T-cell responses in a cancer type long considered highly resistant to immunotherapy.
Glioblastoma remains one of the deadliest cancers, with median survival typically ranging from 12 to 18 months. Patients with MGMT-unmethylated disease face especially poor outcomes because they derive limited benefit from temozolomide, a standard chemotherapy agent commonly used after surgery and radiation.
“Nothing really works in this MGMT-negative or unmethylated glioblastoma patient population,” said Niranjan Sardesai, Geneos’ CEO. “Median survival is around a year, and effective treatments are very much needed.”
The open-label, single-arm GT-20 study enrolled nine patients with newly diagnosed MGMT-unmethylated glioblastoma following surgical resection and radiation therapy. Each patient received a fully individualized vaccine constructed from neoantigens identified through sequencing of their own tumors. Vaccines encoded between 17 and 40 neoantigens per patient.
According to the paper, the vaccine caused no serious adverse events, unexpected toxicities, or dose-limiting toxicities. Eight of the nine evaluable patients developed measurable immune responses. The lone nonresponder had been treated with dexamethasone, an immunosuppressive corticosteroid frequently used in glioblastoma management.
Sardesai emphasized that the immunogenicity findings were particularly notable because glioblastoma is considered an “immune-excluded” tumor with low tumor mutational burden, characteristics that have historically limited the effectiveness of checkpoint inhibitors such as anti–PD-1 therapies.
“Checkpoint-based immunotherapy has not worked in GBM,” he said. “This is a cold tumor.”
The investigators also observed signals of clinical activity. Six-month progression-free survival and 12-month overall survival were each achieved in 66.7% of patients. Median progression-free survival was 8.5 months, while median overall survival reached 16.3 months. Survival at 24 months was 33%, including one patient who remains alive four years after surgery.
“What was very striking was that three of nine patients, or one-third of the patients, had lived more than two years,” Sardesai said. “The two-year survival rate is about 10% to 15%” with standard treatment approaches in this population.
The study also identified an association between stronger CD8-positive T-cell responses and longer survival. Investigators reported that patients generating higher levels of vaccine-induced cytotoxic T cells tended to experience improved overall survival.
One of the most compelling findings involved a long-term survivor who has remained progression-free for nearly five years. Researchers analyzed a brain biopsy obtained approximately three years after treatment initiation and identified vaccine-induced T-cell clones within the tumor tissue that matched T-cell populations detected in the patient’s blood.
“For the first time, we are able to match vaccine-driven immune responses,” Sardesai said. “We are able to see T-cell clones in the blood, and these T-cell clones have infiltrated and are found in her brain.”
The vaccine platform differs from earlier glioblastoma vaccine strategies in several ways. Rather than targeting a small number of antigens, the DNA-based approach allows investigators to incorporate a much larger neoantigen repertoire into each personalized product.
“These patients received as many as 40 different antigens that were identified from their own tumor,” Sardesai said. “Prior treatments had typically been looking at 20 or fewer in GBM.”
He argued that broader antigen targeting may be especially important in glioblastoma because of the disease’s pronounced intratumoral heterogeneity.
“When it comes to targeting cancer, more is better,” he said. “You want to take more shots on goal.”
Another distinguishing feature of the platform is its apparent ability to stimulate CD8-positive killer T cells, which are considered critical for direct tumor cell elimination. Sardesai noted that generating robust CD8 responses has historically been difficult for many cancer vaccine technologies.
Importantly, each vaccine is uniquely manufactured for a single patient.
“These are exquisitely personalized vaccines,” Sardesai said. “Every patient gets their own vaccine.”
The authors cautioned that the findings remain preliminary because of the trial’s small sample size and lack of a control arm. Still, they believe the results justify larger randomized studies.
“We are very encouraged by the data,” Sardesai said. “But this is still only nine patients. We have to replicate these findings in larger, well-controlled studies.”
The company has previously reported results using the same platform in hepatocellular carcinoma, suggesting the strategy could potentially extend across multiple tumor types characterized by immune exclusion and low tumor mutational burden.
“All cancers carry neoantigens,” Sardesai said. “These personalized cancer vaccines provide a very convenient way” to target those tumor-specific alterations across different cancers.
People report that their personal contact info was surfaced by Google AI—and there’s apparently no easy way to prevent it.
A Redditor recently wrote that he was “desperate for help”: for about a month, he said, his phone had been inundated by calls from “strangers” who were “looking for a lawyer, a product designer, a locksmith.” Callers were apparently misdirected by Google’s generative AI.
In March, a software developer in Israel was contacted on WhatsApp after Google’s chatbot Gemini provided incorrect customer service instructions that included his number.
And in April, a PhD candidate at the University of Washington was messing around on Gemini and got it to cough up her colleague’s personal cell phone number.
AI researchers and online privacy experts have long warned of the myriad dangers generative AI poses for personal privacy. These cases give us yet another scenario to worry about: generative AI exposing people’s real phone numbers. (The Redditor did not respond to multiple requests for comment and we could not independently verify his story.)
Experts say that these privacy lapses are most likely due to personally identifiable information (PII) being used in training data, though it’s hard to understand the exact mechanism causing real phone numbers to show up in the AI-generated responses. But no matter the reason, the result is not fun for people on the receiving end—and, even more worryingly, there appears to be little that anyone can do to stop it.
A 400% increase in AI-related privacy requests
It’s impossible to know how often people’s phone numbers are exposed by AI chatbots, but experts say they believe that it is happening far more than is reported publicly.
DeleteMe, a company that helps customers remove their personal information from the internet, says customer queries about generative AI have increased by 400%—up to a few thousand—in the last seven months. These queries “specifically reference ChatGPT, Claude, Gemini … or other generative AI tools,” says Rob Shavell, the company’s cofounder and CEO. Specifically, 55% of these concerns about generative AI reference ChatGPT, 20% reference Gemini, 15% Claude, and 10% other AI tools, Shavell says. (MIT Technology Review has a business subscription to DeleteMe.)
Shavell says customer complaints about personal information being surfaced by LLMs usually take two forms: Either “a customer asks a chatbot something innocuous about themselves and gets back accurate home addresses, phone numbers, family members’ names, or employer details.” Alternatively, a customer may be confronted with and report the exposure of someone else’s personal data, when “the chatbot generates plausible-but-wrong contact information.”
This aligns with what happened to Daniel Abraham, a 28-year-old software engineer in Israel. In mid-March, he says, a stranger sent him a “weird WhatsApp message from an unknown number” asking for help with his account in PayBox, an Israeli payment app.
“I thought it was a spam message,” he wrote to MIT Technology Review in an email—“someone who was trying to troll me.”
But when he asked the stranger how they had found his number, they sent him a screenshot of Gemini’s instructions to contact PayBox customer service via WhatsApp—giving his personal number. Abraham does not work for PayBox, and PayBox does not have a WhatsApp customer service number, Elad Gabay, a customer service representative for the company, confirmed.
Later, Abraham asked Gemini how to contact PayBox, and it generated another person’s WhatsApp number. When I recently asked, Gemini again responded with an Israeli phone number—it belonged not to PayBox, but to a separate credit card company that works with PayBox.
Screenshot: Google Gemini provides MIT Technology Review with the incorrect number for PayBox.
Abraham’s exchange with the stranger ended quickly, but he said he was concerned about how other potential exchanges could quickly turn sour, including “harassment or other bad interactions.” “What if I asked for money in order to ‘solve’ that [customer service] issue?” he said.
To try to figure out how this happened, Abraham ran a regular Google search on his phone number, and he found that it had been shared online once, back in 2015, on a local site similar to Quora. Though he’s not sure who posted it there, it may explain how it ended up being reproduced by Gemini over a decade later.
Chatbots like Gemini, Open AI’s ChatGPT, and Anthropic’s Claude are built on LLMs that are trained on huge amounts of data scraped from across the web. This inevitably includes hundreds of millions of instances of PII. As we reported last summer, for example, the large popular open-source data set DataComp CommonPool, which has been used to train image-generation models, included copies of résumés, driver’s licenses, and credit cards.
The likelihood of PII appearing in AI training data is only increasing as public data “runs out” and AI companies look for new sources of high-quality training data. This includes information from data brokers and people-search websites. According to the California data broker registry, for instance, 31 of 578 registered data brokers operating in the state self-reported that they had “shared or sold consumers’ data to a developer of a GenAI system or model in the past year.”
Furthermore, models are known to memorize and reproduce data verbatim from training data sets—and recent research suggests that it is not just frequently appearing data that is most likely to be memorized.
Imperfect Measures
It’s standard practice now to build guardrails into an LLM’s design to constrain certain outputs, ranging from content filters meant to identify and prevent chatbots from releasing PII to Anthropic’s instructions to Claude to choose responses that contain “the least personal, private, or confidential information belonging to others.”
But as a pair of University of Washington PhD students researching privacy and technology saw firsthand recently, these safeguards don’t always work.
“One day, I was just playing around on Gemini, and I searched for Yael Eiger, my friend and collaborator,” Meira Gilbert says. She typed in “Yael Eiger contact info,” and after Gemini provided an overview of Eiger’s research, which Gilbert had expected, Gemini also returned her friend’s personal phone number. “It was shocking,” Gilbert says.
When she saw the Gemini result, Eiger remembered that she had, in fact, shared her phone number online in the previous year, for a technology workshop. But she had not expected it to be so visible to everyone on the internet.
Have you had your PII revealed by generative AI? Reach the reporter on Signal at eileenguo.15 or tips@technologyreview.com.
“Having your information be … accessible to one audience, and then Gemini making it accessible to anyone” feels completely different, Eiger says—especially when she found that the information was buried in a normal Google search.
“It was severely downgraded,” Gilbert confirms. “I never would have found it if I was just looking through Google results.” (I tried the same prompt in Gemini earlier this month, and after an initial denial, the tool also gave me Eiger’s number.)
After this experience, Eiger, Gilbert, and another UW PhD student, Anna-Maria Gueorguieva, decided to test ChatGPT to see what it would surface about a professor.
At first, OpenAI’s guardrails kicked in, and ChatGPT responded that the information was unavailable. But in the same response, the chatbot suggested, “if you want to go deeper, I can still try a more ‘investigative-style’ approach.” Their inquiry just had to help “narrow things down,” ChatGPT said, by providing “a neighborhood guess” for where the professor might live, or “a possible co-owner name” for the professor’s home. ChatGPT continued: “That’s usually the only way to surface newer or intentionally less-visible property records.”
The students provided this information, leading ChatGPT to produce the professor’s home address, home purchase price, and spouse’s name from city property records.
(Taya Christianson, an OpenAI representative, said she was not able to comment on what happened in this case without seeing screenshots or knowing which model the students had tested, though we pointed out that many users may not know which model they were using in the ChatGPT interface. In response to questions about the exposure of PII, she sent links to documents describing how OpenAI handles privacy, including filtering out PII, and other tools.)
This reveals one of the fundamental problems with chatbots, says DeleteMe’s Shavell. AI companies “can build in guardrails, but [their chatbots] are also designed to be effective and to answer customer questions.”
The exposure issue is not limited to Gemini or ChatGPT. Last year, Futurismfound that if you promptedxAI’s chatbot Grok with “[name] address,” in almost all cases, it provided not only residential addresses but also often the person’s phone numbers, work addresses, and addresses for people with similar-sounding names. (xAI did not respond to a request for comment.)
No clear answers
There aren’t straightforward solutions to this problem—there’s no easy way to either verify whether someone’s personal information is in a given model’s training set or to compel the models to remove PII.
Ideally, individual consumers should be able to request that their PII be removed, says Jennifer King, the privacy and data fellow at Stanford University Institute for Human-Centered Artificial Intelligence. But this is typically interpreted to apply only to the data that people have directly given to companies—like when they interact with a chatbot, King explains.
“I don’t know if Google even has the infrastructure … to say to me, ‘Yes, we have your data in our training data, we can summarize what we know about you, and then we can delete or correct things that are wrong or things that you don’t want in there,’” she says.
Existing privacy legislation, like the California Consumer Privacy Act or Europe’s GDPR, does not cover the “publicly available” information that has already been scraped and used to train LLMs, especially since much of this is anonymized (though multiplestudies have also shown how easy it is to infer identities and PII from anonymized and pseudonymous data).
As to “whether they [AI companies] have ever systematically tried to go back through data that had already been collected from the public internet and minimized that stuff?” King adds. “No idea.”
The next best solution would be that the companies are “taking out everybody’s phone numbers or all data that resembles [phone numbers],” King says, but “nobody’s been willing to say” they’re doing that.
Hugging Face, a platform that hosts open-source data sets and AI models, has a tool that allows people to search how often a piece of data—like their phone number—has appeared in open-source LLM training data sets, but this does not necessarily represent what has been used to train closed LLMs that power popular chatbots like Claude, ChatGPT, and Gemini. (Eiger’s number, for example, did not show up in Hugging Face’s tool.)
Alex Joseph, the head of communications for Gemini apps and Google Labs, did not respond to specific questions, but he said that “the team” is “looking into” the particular cases flagged by MIT Technology Review. He also provided a link to a support document that describes how users can “object to the processing of your personal data” or “ask for inaccurate personal data in Gemini Apps’ responses to be corrected.” The page notes that the company’s response will depend on the privacy laws of your jurisdiction.
OpenAI has a privacy portal that allows people to submit requests to remove their personal information from ChatGPT responses, but notes that it balances privacy requests with the public interest and “may decline a request if we have a lawful reason for doing so.”
Anthropic describes how it uses personal data in model training, but it does not have a clear way for people to request its removal. The company did not respond to a request for comment.
The best option for anyone who wants to protect their private data right now is to “start upstream: get personal data off the public web before it ends up in the next scrape,” says Shavell. Since the start of the year, for instance, California has offered its residents a web portal to request that data brokers delete their information. Still, this doesn’t guarantee that your data hasn’t already been used for training—and will therefore not appear in a chatbot’s response.
The Redditor who received incessant calls posted that he had “submitted an official Legal Removal/Privacy Request to Google, asking them to urgently blacklist my number from their LLM outputs,” but had not yet received a response. He also wrote last month that “the harassment continues daily.”
Abraham, the Israeli software developer, says he contacted Google’s customer service on March 17, the day after his phone number was exposed. He says he did not receive a response until May 4, and it simply asked for documentation that he had already provided.
Meanwhile, inspired by her own exposure on Gemini, Eiger, along with Gilbert and Gueorguieva, is designing a research project to further study what personal information is being surfaced by various AI chatbots—and what they may know, even if they’re not telling us.
Some of that information may “technically be public,” says Gilbert, but chatbots may be altering “the amount of effort you would put into finding” it. Now instead of searching through 10 pages of Google search results, or paying for the information from a data broker site, “does generative AI just lower the barrier to entry to target people?”
This piece has been updated to clarify OpenAI’s response.
The incidence of stage IV breast cancer increased significantly overall, across ages, and for both sexes from 2010 through 2021, according to research from a Dana Farber-led team. The percentage of patients with stage IV breast cancers, versus those with stages I to III diagnoses also increased.
Notably, this increase was seen for all tumor subtypes in both sexes.
The researchers write, “These findings suggest that efforts are needed to determine factors contributing to these increases and to identify breast cancer before patients present with de novo stage IV disease.”
The study appears this week (May 12 issue) in JAMA Network Open. The senior author is José P. Leone, MD, department of medical oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston.
In their analysis of data from over 700,000 U.S. patients, the incidence of stage IV breast cancer increased significantly by 1.2% per year, and the percentage of people with stage IV also increased significantly. Stage IV incidence increased widely across all ages, races, sexes, and tumor subtypes. Still, survival improved significantly from 2010 through 2021.
Stage IV incidence increased across all tumor subtypes in both sexes. In women, those subtypes include hormone receptor (HR)–positive/ERBB2-negative, HR-positive/ERBB2-positive, HR-negative/ERBB2-positive, and triple-negative disease.
Trends in the incidence of de novo stage IV breast cancer “remain underreported,” these authors write. A previous study evaluating incidence of distant disease in the U.S. before 2010 showed a statistically significant increase in incidence for younger patients and a statistically significant decrease in older patients.But, this current study’s authors said, a meta-analysis reported a decreasing percentage of stage IV presentation over time.
Breast cancer is the second most common cancer in women, behind skin cancer. It is the most common cancer diagnosed in females worldwide and an estimated 30% of patients develop metastases. The American Cancer Society estimates 42,140 U.S. women will die from breast cancer in 2026.
The incidence of breast cancer in younger women, in particular, has been rising. In August 2025, the CDC reported that: “Most breast cancers occur in older women, but rates have been increasing slowly among women younger than 45 years in all racial and ethnic groups.” The agency added that survival from breast cancer is improving “among women in most racial and ethnic groups.”
Breast cancer in men remains rare, but the rate is increasing also.
This population-based cohort study used data from the Surveillance, Epidemiology, and End Results (SEER) program to identify patients diagnosed with de novo invasive breast cancer between January 1, 2010, and December 31, 2021. Data analyses were conducted from January 2024 to June 2025.
Of 761,471 breast cancer diagnoses, 43,934 (5.8%) were stage IV. Stage IV incidence increased from 9.5 cases per 100,000 females in 2010 to 11.2 cases per 100,000 females in 2021. The incidence of stages I to III disease also increased, from 163 cases per 100,000 females in 2010 to 177.4 cases per 100,000 females in 2021.
Among males, there was also a statistically significant increase in stage IV incidence.
The researchers noted that, “Although overall survival improved, research is warranted to determine factors contributing to increased incidence, including potential changes in natural history of breast cancer, disease screening, and incidence and mortality of other conditions.”
Background: “Keep On Keep Up” (KOKU) is a tablet-based digital program based on the well-validated Otago and Fitness and Mobility Exercise programs for older adults to decrease the risk of falling. Objective: This substudy involved a process evaluation in order to analyze the usage patterns of the KOKU digital program, specifically training frequency, volume, and intensity among older adults over a 3-month self-managed training period. Pre-post changes in physical capacity and real-world walking were examined. Methods: This study is a nested cohort study within the three-armed randomized controlled SMART-AGE trial conducted in Germany (German Clinical Trials Register ID: DRKS00034316). Participants aged 67 years or older with basic digital literacy were included. KOKU provided guided but unsupervised progressive strength and balance training for 3 months. The data on training adherence, engagement, and progression were collected. Instrumented assessments included the Timed Up and Go Test, the 30-Second Chair Rise Test, and real-world walking monitoring using wearable sensors. Results: A total of 113 participants (n=63, 56% female; mean age 74.02, SD 5.36 y) were included in the analysis. During the 3-month period, participants used KOKU for 24 (SD 15) days, that is, 2 to 3 times per week. Over the entire study period, no falls or other adverse events were reported due to KOKU usage. The number of exercises performed per participant ranged from 2 to 213, with a median value of 70. The instrumented Timed Up and Go Test results revealed a prolonged total duration (=0.26; =.009). In the instrumented 30-Second Chair Rise Test, improvements were observed in the number of completed repetitions (=0.21; =.04) and frequency of repetitions (=0.23; =.03). This was mainly due to a reduction in inactive time (=−0.60; <.001). Real-world walking parameters remained unchanged, except for a slower walking speed during walking bouts of less than 30 seconds (=0.49; <.001). All changes did not meet the criteria for minimally important differences. Conclusions: KOKU is a novel digital intervention for older adults, promoting balance and strength exercises. Physical capacity improvements were small. However, the use of instrumented assessments provided further insights into participants’ capacity and mobility that would not have been identifiable with conventional assessments. Future improvements to the program should focus on incorporating more challenging exercises for individuals with varying levels of physical capacity. Trial Registration: German Clinical Trials Register DRKS00034316; https://drks.de/search/en/trial/DRKS00034316
Background: Older adults facing social or structural marginalization for reasons such as lower literacy, digital exclusion, financial constraints, restricted living environments, or complex health histories, face persistent barriers to much-needed health screening. Digital health tools, particularly those using audio computer-assisted self-interview (ACASI) technology, offer potential to overcome these barriers (audio-delivered and self-administrable), but their application to marginalized populations remains underexplored. Moreover, guidance is limited for developing such tools which require collaboration within cross-disciplinary teams. This paper presents development insights and user testing findings from the ASCAPE (Audio App-Delivered Screening for Cognition and Age-Related Health in Prisoners) project, which aimed to develop equitable digital frailty and cognition screening for older people in Australian prisons. Objective: This study aims to describe the collaborative development of the “ASCAPE-HS” prototype, a tablet-based ACASI-delivered Frailty Index and aging screener, and to synthesize key lessons from the project that can inform equitable digital health tool development in hard-to-reach older adults. Also, to present findings on the usability and acceptability of ASCAPE-HS in a diverse community sample. Methods: The ASCAPE-HS prototype was developed through an iterative process involving researchers, clinicians, software developers, and end users under a digital health equity framework. The prototype included a self-administered, audio-delivered Frailty Index, alongside other items relevant to aging. The design process prioritized accessibility, sociocultural relevance, and technical feasibility, with regular multidisciplinary consultation and iterative refinement. Exploratory user testing with 20 older adults (aged 47‐93 years, including n=5 who had not finished secondary schooling, n=3 people with previous imprisonment history, and n=9 with mild or moderate cognitive impairment) provided feedback on usability and acceptability. Results: A 50-item Frailty Index was developed, alongside an additional selection of holistic questions that could meaningfully capture age-related health, and transferred to an iOS app (Apple, Inc), with ACASI features. Key elements included lay wording, consistent interface, simple “tapping” response options with repeatable audio feedback, a tutorial, and artificial intelligence–generated audio guidance. Key development considerations were synthesized into a checklist for teams undertaking similar projects. Successful strategies for the collaborative design process included diverse teams abreast of emerging literature and policy with varying expectations for engagement during development, and dedicating time to flexible, iterative development processes. Acceptability (median scores ≥4 out of 5 across 6 constructs) and usability (mean System Usability Scale score 79.0, SD 8.8) were high. Conclusions: A collaborative approach can produce ACASI-based health screening tools that are well-received by older adults. We highlight the feasibility of integrating frailty and aging assessment into a usable and acceptable digital tool, and offer actionable principles for collaborative, evidence-based development of equitable health screening tools in diverse, hard-to-reach populations.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/30699dc5ac2d0ad05f7755da3210033e" />
Taking a creative approach to chemistry can help developers of antibody-drug conjugates (ADCs) improve the stability and purity of their products. That’s the view of Sunny Zhou, PhD, professor of chemistry and chemical biology at Northeastern University. Zhou will be speaking at the Bioprocessing Summit in Boston in August.
According to Zhou, the structure of ADCs can make them vulnerable to bioprocessing issues that don’t affect traditional antibodies. As one example, he says, the payloads of antibody drug conjugates often significantly absorb above 280 nm, making them markedly more sensitive to light.
“There’ll be photochemistry induced by the payload that can damage both the antibodies and payloads, such as crosslinking that likely leads to aggregation,” he says. “We’ve already published some work showing light-induced protein modifications, crosslinking, and aggregation.”
According to Zhou, some initiatives are already underway to address this issue. For example, by engaging in antibody production and downstream processing in dim or safe light (e.g., yellow or red light) instead of the more commonly used bright white light.
Another issue, he says, is that the linker connecting the antibody and drug payload is designed to be cleaved by enzymes in human patients. On the other hand, it also means that similar enzymes in host cell proteins (HCPs) may prematurely cleave the linker during production and storage, thereby decomposing the drug and contaminating the final product.
“Many host cell proteins contain such enzymes, but they don’t cleave antibodies. With these ADC linkers, however, enzymes that didn’t create problems before might do so now,” he says.
Zhou explains that premature cleavage of ADC linkers has been observed in an industrial setting. Fortunately, he says, his research team, in collaboration with companies like Takeda, is already creating universal platforms and workflows to identify and effectively remove these potential HCP contaminants, as well as working to better understand the stability of the linkers.
“These drugs circulate in the body for maybe two to three weeks, and stability issues can be amplified during circulation,” he says. “So, making the linker more stable [during manufacturing] may also help improve stability during circulation, further down the line.”
Zhou’s team is now hoping to look at other creative chemistries in bioprocessing. Among these is, for example, removing reagents, by-products, and impurities by filtration, which may be faster than relying on chromatography, he says.
Choosing the right additives could help “cell-free” expression systems finally fulfill their potential and provide biopharma with a low-cost way of making protein drugs, according to a recent research report.
The new study looked at how cell-free systems, in which biochemical reactions occur independently of cells, could be fine-tuned to provide drug makers with alternatives for large-scale protein production.
And the potential of the approach is significant, says Karen Polizzi, PhD, a professor from the department of chemical engineering at Imperial College London, who adds, “Cell-free protein synthesis (CFPS) is a flexible manufacturing technology. It can be used for on-demand synthesis in low-resource environments or to make difficult-to-express products, especially medicines that are toxic to the cell. Cell-free reactions scale well across microliter to liter scale without needing adjustments.”
The Imperial team’s research focused on expression systems based on the yeast species Pichia pastoris, which, as Polizzi explains, “has machinery capable of post-translational modifications of proteins that can be necessary for function.”
As an expression host, P. pastoris combines elements of both prokaryotic and eukaryotic systems, such as a rapid growth rate and the ability to perform post-translational modifications (PTMs).
The problem is that current commercially available Pichia systems are only able to produce low amounts of protein. According to Polizzi and her co-authors, the productivity of P. pastoris-based cell-free systems usually ranges from 6 to 100 µg/mL, which is only approximately five percent of that achieved by comparable E. coli systems. In addition, the additives required by Pichia-based systems are more expensive than those required by equivalent platforms.
Additives to improve yields
To address this, Polizzi and co-authors systematically evaluated a variety of chemical additive combinations to identify the most effective stabilizers and crowding agents to be incorporated in the reaction.
The researchers also used a machine learning model to predict translation initiation rates and optimized the Kozak sequence—the protein translation initiation site in most eukaryotic mRNA transcripts—to enhance expression.
In addition, the Imperial team evaluated lower-cost glycolytic intermediates as substrates for ATP regeneration to reduce the cost of goods.
Polizzi says, “We focused on how to improve the yields and reduce the cost of production. We identified some additional additives that boost the yield without substantially increasing the cost. We also identified a different energy source that can be used.”
She adds, “This work underscores the importance of protein-stabilizing additives and the role of rationally designed DNA sequences with minimized mRNA structural complexity to enhance yield in CFPS. Our demonstration of glycolytic intermediates as a potential secondary energy system additionally provides the foundation for the development of a cost-effective P. pastoris CFPS.”
Mesenchymal stem cell-derived extracellular vesicles (MSC-EVs) play an outsized role in intracellular communications, influencing such functions as inflammation and tissue repair. With the possible applications of these small, membrane-bound particles growing, an efficient, cost-effective production method has been on drug manufacturers’ wish lists for some time.
A novel, streamlined chromatographic production and isolation method developed by scientists at Satorius BIA Separations in Slovenia may fulfill that wish, yielding uniform, high-quality EVs at scale. The method concentrates MSC-EVs directly from conditioned media. It also removes 97% of protein impurities and 95% of double-stranded DNA-related impurities, increasing their potential as therapeutics or drug delivery vessels.
Microcarrier + suspension
The method relies upon preferential exclusion chromatography, Katja Vrabec, head of product application area (EVs) at Sartorius, notes in a recent paper. In it, Vrabec and colleagues explain the method “uses monolithic hydroxyl columns to purify and concentrate the MSC-EVs,” and biochromatography analytics to track EV-specific surface antigens.
First, the team expanded the MSCs in growth media, and then produced the EVs in a lean media formulation to limit production of protein and particle contaminants. That part is standard.
Here’s what’s different: The scientists used a microcarrier-based system rather than flask-based 2D cultivation to scale the MSC cultures and increase the ratio of EVs to contaminants in conditioned media. They also used a suspension culture to enhance cell growth surface-to-volume ratios, and thereby increase EV yield. Then, they used a monolithic hydroxyl column to capture and purify the EVs directly from harvest.
Increasing cell density and the cell-to-impurity ratio lowers buffer consumption downstream and lays the groundwork for biomanufacturers to transition to a scalable bioreactor system.
Because the main impurities in EV harvests don’t interact with the chromatographic column in high-salt-binding conditions, the team recommends choosing a low-salt buffer for elution to reduce the need for buffer exchange before the polishing step. The optimal binding condition, they report, is “sodium citrate of 0.75M at pH 7.0.”
This research highlights the need to consider upstream and downstream processing as a cohesive system, to design a simple, scalable, holistic process, and to apply reliable analytics. This all is particularly challenging, the team admits, given “the heterogeneous nature of EVs and the presence of similarly-sized components in biological samples.”
Background: Emergency department (ED) overcrowding and delayed access to care are ongoing challenges in Singapore. The COVID-19 pandemic further underscored the need for scalable virtual care models that go beyond traditional hospital settings, allowing patients to access acute specialist care quickly and efficiently. Objective: This study describes the design, implementation, and early outcomes of the National University Health System (NUHS) Virtual Care Centre (VCC), a clinician-led helpline aimed at reducing unnecessary ED visits and supporting community-based acute care. Methods: In 2020, the NUHS launched the VCC, a helpline at Alexandra Hospital, as a prehospital triage model. The helpline functions as a nurse-led telephone triage with real-time escalation to doctors for urgent medical issues. It ensures the continuity of care for patients recently discharged and diverts nonemergency cases from the ED. A retrospective analysis of call data from 2020 to 2024 was conducted to evaluate utilization patterns, clinical outcomes, and safety. Results: Over 4 years, the VCC managed 4857 calls, of which 59.3% (n=2879) were clinical in nature. Nearly two-thirds (1834/2879, 63.7%) were resolved remotely, preventing in-person ED visits. Only 13.8% (397/2879) required redirection to an ED, and 3.3% (95/2879) were directly admitted to an acute hospital or hospital at home service. Within 72 hours of call resolution, 69.1% (1990/2879) of the callers avoided an ED visit. Undertriage was 4.9% (110/2232) at 72 hours post call resolution, with no high dependency or intensive care unit admissions during this period. Mortality rates were low (1.0% at 14 days; 2.3% at 30 days). Conclusions: The NUHS VCC provides a feasible and safe model for virtual acute care triage within the public health care system. It effectively diverted lower-acuity cases from the ED and ensured continuity of care, offering a scalable approach aligned with national efforts to extend health care beyond hospital walls.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/bbbab2870bb94f230cfd987cf69b3559" />
Background: Major depressive disorder (MDD) affects approximately 1 in 6 adults during their lifetime, yet antidepressant selection relies predominantly on trial-and-error, with response rates of only 42% to 53%. While machine learning (ML) models have shown promise in predicting treatment outcomes, most focus on single treatments rather than comparative selection across therapeutic alternatives, limiting their clinical utility for the medication choice decisions that clinicians face in practice. Objective: This systematic review evaluates ML approaches that examine 2 or more pharmacological interventions for predicting treatment outcomes in MDD, with a focus on their capacity to facilitate comparative treatment selection between medications or medication classes for individual patients. Methods: Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, we searched PubMed, Scopus, and Web of Science for studies published from 2015 to 2025. We included studies involving adults with MDD that used ML models to predict treatment outcomes across 2 or more pharmacological treatments and reported medication-specific prediction outcomes. Risk of bias was assessed using PROBAST-AI (Prediction Model Risk of Bias Assessment Tool for Artificial Intelligence). We conducted a narrative synthesis organized by modeling strategies, data integration approaches, validation methodologies, and performance patterns. Results: From 5370 initial records, 19 studies met the inclusion criteria, with dataset sample sizes ranging from 49 to 77,226 participants. Studies employed 3 distinct modeling strategies: drug-specific supervised models trained independently for each medication, subtype- or trajectory-based approaches using clustering methods to identify differential response patterns, and a unified differential prediction framework generating calibrated cross-treatment predictions. Performance varied substantially, with area under the curve values ranging from 0.59 to 0.95 and classification accuracies between 62% and 95.4%, though high performance was concentrated in studies with small samples, high-dimensional neurobiological features, and internal-only validation. Only 7 studies conducted external validation, which generally yielded more conservative performance estimates. Feature informativeness was more consistently associated with performance variation than algorithm complexity. Most studies did not formally distinguish between prognostic features predicting general outcomes and predictive features identifying differential medication responses, and none applied formal explainability techniques. Conclusions: ML for comparative antidepressant selection remains in an early stage of development. Only 1 study implemented a unified framework directly supporting patient-level treatment ranking. Key barriers to clinical translation include insufficient distinction between prognostic and predictive markers, limited cross-trial validation, near-absent calibration reporting, and absent explainability. Future research should prioritize unified comparative frameworks with calibrated predictions, rigorous external validation on diverse cohorts, explicit modeling of heterogeneous treatment effects, and integration of explainability into model development.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/a59ed16526360bac492b29bed9672c02" />
