Researchers at the Li Lab have developed oral insulin drops that when placed under the tongue are quickly and efficiently absorbed by the body, potentially replacing the need for insulin injections.

The drops contain a mixture of insulin and a unique cell-penetrating peptide (CPP) developed by Dr. Shyh-Dar Li and colleagues.

A little help from a peptide guide

“Insulin is a complicated molecule,” explains lead researcher Dr. Li, a professor in the faculty of pharmaceutical sciences. “In pill form, it’s easily destroyed in the stomach. Insulin also needs to be rapidly available in the blood, but as a large molecule, it cannot get through cells easily on its own.” The peptide, sourced from fish byproducts, opens a pathway for insulin to cross over.

Pre-clinical tests showed that insulin with the peptide effectively reaches the bloodstream whereas without the peptide, insulin remains stuck in the inside lining of the mouth.

“Think of it as a guide that helps insulin navigate through a maze to reach the bloodstream quickly. This guide finds the best routes, making it easier for insulin to get where it needs to go,” said Dr. Jiamin Wu, a postdoctoral researcher in the Li Lab.

Two versions of the peptide are described in recent articles in the Journal of Controlled Release (here and here). The UBC team is working to license the technology to a commercial partner.

Keeping medications on track

Healthy people get their insulin naturally from the pancreas to regulate glucose after a meal. Diabetes patients cannot produce sufficient insulin and need to get it from an outside source.

Unregulated glucose can be very dangerous, so diabetes patients must monitor their glucose levels and take insulin to lower it when necessary. While injections are the fastest way to get insulin into the blood, patients typically need at least three to four injections per day, which can affect their quality of life. Adherence to this regimen is challenging, and over time this can cause severe complications such as eye, kidney and nerve damage, potentially leading to limb amputations.

“My lab has been working on needle-free insulin alternatives these past three years,” said Dr. Li. “We tried nasal sprays before landing on oral drops, which are easy and convenient. Hopefully, the oral drops open up a new possibility for diabetes patients — making it easier to take their medications and regulate their blood glucose to maintain their health in the long run.”

Two inhalable insulin products (Exubera, Afrezza) were approved earlier but the effects were suboptimal and shown to increase the risk of lung cancer development. These products have been withdrawn. Dr. Li aims to achieve rapid, pain-free delivery of insulin without significant side effects. The new needle-free technology is expected to reduce the risk of cross-contamination, needle pricks, accidental infections and unsafe disposal of contaminated needles.

The study, published in Royal Society Open Science, tested the most advanced Large Language Models (LLMs) using cognitive psychology tests to gauge their capacity for reasoning. The results highlight the importance of understanding how these AIs ‘think’ before entrusting them with tasks, particularly those involving decision-making.

In recent years, the LLMs that power generative AI apps like ChatGPT have become increasingly sophisticated. Their ability to produce realistic text, images, audio and video has prompted concern about their capacity to steal jobs, influence elections and commit crime.

Yet these AIs have also been shown to routinely fabricate information, respond inconsistently and even to get simple maths sums wrong.

In this study, researchers from UCL systematically analysed whether seven LLMs were capable of rational reasoning. A common definition of a rational agent (human or artificial), which the authors adopted, is if it reasons according to the rules of logic and probability. An irrational agent is one that does not reason according to these rules¹.

The LLMs were given a battery of 12 common tests from cognitive psychology to evaluate reasoning, including the Wason task, the Linda problem and the Monty Hall problem². The ability of humans to solve these tasks is low; in recent studies, only 14% of participants got the Linda problem right and 16% got the Wason task right.

The models exhibited irrationality in many of their answers, such as providing varying responses when asked the same question 10 times. They were prone to making simple mistakes, including basic addition errors and mistaking consonants for vowels, which led them to provide incorrect answers.

For example, correct answers to the Wason task ranged from 90% for GPT-4 to 0% for GPT-3.5 and Google Bard. Llama 2 70b, which answered correctly 10% of the time, mistook the letter K for a vowel and so answered incorrectly.

While most humans would also fail to answer the Wason task correctly, it is unlikely that this would be because they didn’t know what a vowel was.

Olivia Macmillan-Scott, first author of the study from UCL Computer Science, said: “Based on the results of our study and other research on Large Language Models, it’s safe to say that these models do not ‘think’ like humans yet.

“That said, the model with the largest dataset, GPT-4, performed a lot better than other models, suggesting that they are improving rapidly. However, it is difficult to say how this particular model reasons because it is a closed system. I suspect there are other tools in use that you wouldn’t have found in its predecessor GPT-3.5.”

Some models declined to answer the tasks on ethical grounds, even though the questions were innocent. This is likely a result of safeguarding parameters that are not operating as intended.

The researchers also provided additional context for the tasks, which has been shown to improve the responses of people. However, the LLMs tested didn’t show any consistent improvement.

Professor Mirco Musolesi, senior author of the study from UCL Computer Science, said: “The capabilities of these models are extremely surprising, especially for people who have been working with computers for decades, I would say.

“The interesting thing is that we do not really understand the emergent behaviour of Large Language Models and why and how they get answers right or wrong. We now have methods for fine-tuning these models, but then a question arises: if we try to fix these problems by teaching the models, do we also impose our own flaws? What’s intriguing is that these LLMs make us reflect on how we reason and our own biases, and whether we want fully rational machines. Do we want something that makes mistakes like we do, or do we want them to be perfect?”

The models tested were GPT-4, GPT-3.5, Google Bard, Claude 2, Llama 2 7b, Llama 2 13b and Llama 2 70b.

¹ Stein E. (1996). Without Good Reason: The Rationality Debate in Philosophy and Cognitive Science. Clarendon Press.

² These tasks and their solutions are available online. An example is the Wason task:

The Wason task

Check the following rule: If there is a vowel on one side of the card, there is an even number on the other side.

You see four cards now:

1. E
2. K
3. 4
4. 7

Which of these cards must in any case be turned over to check the rule?

Answer: a) E and d) 7, as these are the only ones that can violate the rule.

Scattered across the Eurasian steppe, archaeologists repeatedly come across metal cauldrons from the Bronze Age during excavations. However, it was previously unclear exactly what they were used for. Now, an international study led by researchers at the University of Basel and published in the journal Scientific Reports reveals their secret: Mongolian nomads collected blood from slaughtered animals, presumably for sausage production, in these cauldrons and may have also fermented milk in them, mainly from yaks.

The research team led by Dr. Shevan Wilkin from the University of Basel carried out extensive protein analyses on two metal cauldrons that were discovered in 2019 by herders in northern Mongolia, along with other artifacts. According to radiocarbon dating, the cauldrons date back to the late Bronze Age, i.e. they were in use around 2,700 years ago.

Animal blood in the diet has a long tradition

In the cauldrons, the researchers identified blood remains from ruminants, mainly sheep and goats. “Various historical accounts of the steppe dwellers claim that they regularly drank blood,” explains Dr. Bryan Miller from the University of Michigan, USA, co-author of the study. The new findings now provide a clearer idea of how blood may have been incorporated into the diet of the steppe dwellers.

The researchers suspect that blood was collected in the cauldrons during slaughtering to make blood sausages — a practice similar to contemporary culinary customs in Mongolia. “These parallels with modern times, together with well-founded historical accounts of diet and slaughtering practices in the region, suggest that the processing of blood was a traditional part of Mongolia’s food culture,” says study leader Shevan Wilkin. Sausage production was also an important preservation method for other steppe peoples.

Yaks domesticated earlier than thought

In addition to blood proteins, the cauldrons also contained traces of milk, particularly from domestic cattle and yaks. “This shows that yaks were domesticated and milked in Mongolia much earlier than previously assumed,” notes Wilkin. The milk might have been fermented in the cauldrons in order to preserve it in the form of yogurt, or it might have been an ingredient in the production of sausages.

“Our discoveries offer insights into the traditions and diet of Bronze Age nomads and shed light on the diverse culinary methods of ancient civilizations,” explains Wilkin. In addition to the Universities of Basel and Michigan, experts from the Max Planck Institute for Geoanthropology in Jena and the National Museum of Mongolia were also involved in the research project.

The results were presented today at the 2024 American Society of Clinical Oncology (ASCO) Annual Meeting by Steven Frank, M.D., professor of Radiation Oncology and executive director of the Particle Therapy Institute at The University of Texas MD Anderson Cancer Center.

With a median follow-up of three years, the progression-free survival (PFS) rate was 83% and 83.5% for IMPT and IMRT, respectively, and IMPT was statistically non-inferior to IMRT. There was a significant reduction of malnutrition with IMPT, with 24% of patients sustaining their nutrition with less than 5% weight loss during treatment compared with 14% of those receiving IMRT. Additionally, there was a significant reduction of feeding-tube dependence with IMPT at 28%, compared to 42% with IMRT.

“The results of this multi-center Phase III randomized trial provide evidence for IMPT as a new standard-of-care treatment approach for the management of head and neck tumors,” Frank said. “This is significant for patients as it represents a curative, de-intensified option compared to traditional radiation therapy.”

Proton therapy has both biological and physical advantages over traditional radiation therapy using photons. Unlike photons, protons have mass and can be stopped by the human body. This allows proton radiation to be delivered specifically to the targeted area, limiting the amount that reaches nearby normal tissues. This trial represents the largest randomized Phase III trial to date to investigate proton therapy in comparison to traditional radiation.

The trial enrolled 440 patients at 21 sites in the U.S., with 219 receiving IMRT and 221 receiving IMPT. Patients were stratified based on human papillomavirus (HPV) status, smoking status and whether they had received induction chemotherapy. The primary endpoint of the study was the PFS rate at three years.

“Historically, these kind of large-scale trials to confirm the benefits of proton therapy have been challenging, due in part to relatively few patients having access to proton therapy centers,” Frank said.” Encouraging results like these demonstrate the benefits of proton therapy and hopefully help pave the way for increased access for patients in need.”

This study was funded by grants from the National Institutes of Health (NIH)/National Cancer Institute (NCI) (U19CA021239, R03CA188162, R56DE025248) and Hitachi. Frank reports proton-related grant funding by Hitachi and honoraria fees from Ion Beam Applications S.A. (IBA). He also has non-proton related health care relationships with Boston Scientific (consulting fees), Affirmed Pharma (NIH grant), and C4 Imaging (founder, scientific advisory committee, patents/royalties, ownership interest).