Prince Abdullah bin Bandar bin Abdulaziz, Minister of the Saudi National Guard, launched in Riyadh, on Monday, the fourth edition of the “Global Biomedical Technology Summit 2026,” which includes dialogue sessions, scientific meetings, and specialized workshops with the participation of about 100 speakers including CEOs, presidents of international universities, experts, and researchers.
The global summit will be held under the patronage of Prince Mohammed bin Salman, Crown Prince, Prime Minister and Chairman of the Board of Directors of the “Riyadh Biotechnology Center” Foundation, during the period from 14 to 16 September, under the slogan “Building the foundations of excellence in medical biotechnology.”
The Minister of the National Guard attended the opening session of the summit, which brought together an elite group of leaders, experts, researchers, innovators and investors in the medical biotechnology sector from various countries of the world.
The summit discusses several topics related to the future of medical biotechnology, most notably national strategies and regulatory frameworks, infrastructure for research, development and innovation, financing and investment, and building human capabilities and competencies, in addition to reviewing the latest developments and innovations in the sector.
The summit will witness an accompanying exhibition in which governmental, academic, health, technical, industrial and investment bodies participate, to showcase the latest products, services and advanced medical technologies.
It also includes meetings that bring together investors and decision-makers, and includes a platform for signing agreements and memorandums of understanding, which enhances opportunities for cooperation and partnerships in the field of medical biotechnology.
It is noteworthy that the summit is being organized by the Ministry of National Guard, represented by Health Affairs, King Saud bin Abdulaziz University for Health Sciences, and the King Abdullah International Center for Medical Research (KIMARC), in partnership with the Ministry of Investment and the Future Investment Initiative Foundation as a strategic partner.
A new study reveals that two genetic modifications can make CAR-T cells more able to penetrate and attack solid tumors, in a step that may help overcome one of the biggest challenges facing this type of immunotherapy.
Experiments on mice showed that disrupting two genes, GNAS and P2RY8, in immune cells enhanced their ability to reach tumors and continue their anti-cancer activity.
These modifications also helped reduce multiple tumors, including lung, pancreas, skin, and some gastrointestinal and uterine cancers.
The results were published in the journal Nature on August 12, 2026 after using the first large-scale platform based on CRISPR technology to study the effect of genetic modifications in human immune cells inside a living organism.
Combating solid tumors
• Why CAR-T in solid tumors? CAR-T treatment depends on taking immune cells called “T cells” from the patient, then modifying them in the laboratory to recognize cancer cells and attack them before returning them to the patient’s body. Chimeric receptors for T cells (CAR-T) are artificial receptors that were specially created in the laboratory, with the aim of enabling T cells to recognize and target specific proteins on the surface of cells (cancer cells, for example).
The treatment has achieved significant success in some blood cancers, such as leukemia and lymphoma.
But its effectiveness is still limited against solid tumors, including lung, pancreatic, ovarian, colon, and breast cancers. This is partly because the environment surrounding a solid tumor is full of signals that inhibit immune cells, and it is difficult for immune cells to reach the tumor and remain active enough to eliminate cancer cells.
• Testing thousands of genes within tumors. Instead of studying modified cells only in laboratory dishes, researchers at the Gladstone Institutes and the University of California, San Francisco, in the United States, led by Dr. Qi Liu, have developed a method that allows testing the effect of large-scale genetic modifications within tumors in mice.
The team tested nearly 20,000 genes to look for modifications that help T cells reach tumors, or maintain their activity after they enter them.
The researchers were able to collect millions of T cells from each tumor, a much larger number than was possible using previous methods, allowing genetic testing to be performed on a large scale within the real environment of the tumor.
Genetic modification
• Two genes remove the “brakes” of immune cells. The P2RY8 gene has emerged as one of the most important genes that influence the ability of T cells to enter the tumor.
The results showed that disabling it led to the accumulation of greater numbers of immune cells inside tumors after removing one of the signals that appears to act as a “brake” that limits the arrival of T cells.
Disabling the GNAS gene had a different effect. T cells became able to produce higher levels of interferon-gamma, an important molecule in the immune response and attacking cancer cells.
GNAS appears to act as a central loop that receives a set of signals that the environment surrounding the tumor uses to weaken immune cells. When it is inactivated, T cells become less responsive to these inhibitory signals, allowing them to maintain their anti-cancer activity.
• Robust results in multiple models. When the GNAS gene was disabled in CAR-T cells, these cells were able to shrink tumors in multiple mouse models of cancer.
The combination of inactivating both genes was more effective. In one lung cancer model, about two-thirds of the mice remained tumor-free at the end of the experiment despite using a very low dose of CAR-T cells, while cells modified in the traditional way did not achieve the same result.
The researchers also tested the modifications using T cells taken from patients with ovarian cancer and melanoma, and found that the genetic modifications retained their ability to enhance the fight against tumors.
Despite the encouraging results, this strategy is still in the pre-clinical stage, and needs more studies before being tested on humans. The researchers followed the mice for more than six months after the tumors disappeared, and observed no long-term side effects. The team believes that one reason for this is that the modifications do not lead to widespread activation of T cells throughout the body, but rather their effects are more pronounced within the environment surrounding the tumor.
Scientists believe that the importance of the study is not limited to the two discovered genes, as the CRISPR platform within the organism allows the search for other genetic modifications that may make cellular treatments more able to reach tumors and resist their defenses, which may pave the way for a new generation of immunotherapies designed more accurately to confront solid cancers.
With the expansion of the development of self-driving cars, the challenge is no longer limited to the ability of artificial intelligence systems to make decisions while driving, but it has also become necessary for humans to understand the reasons behind those decisions, especially when the vehicle behaves in an unexpected way.
These cars rely on deep learning models that act as the “brain” of the vehicle, as they process data received from cameras and sensors to form a perception of the surrounding environment, then determine what the vehicle should do and chart the path it will take. However, these models often act as a “black box,” which makes it difficult to know the real reason behind some of their decisions, especially when an error occurs.
“Conceptual Framing Network”
In this context, researchers from the Massachusetts Institute of Technology (MIT) in the United States, in cooperation with Motional, a company specializing in self-driving vehicles, developed a new method they called the “Concept-Wrapper Network,” which aims to transform the decision-making process within self-driving systems into explanations that are understandable to humans, helping them predict the car’s behavior more accurately. The model was designed to work within the planning system used in the car, without the need to retrain the original model from scratch, or sacrifice its performance.
The idea, according to the team, is to replace a concept classifier with the final stage responsible for evaluating leadership paths in the original model, followed by a new decision-making stage. Thus, the concepts that the system recognizes become an actual part of the decision-making process, not just a description added after it is made. The results were published in the September 2, 2026 issue of the journal “Nature.”
To train the system, the researchers used two data sets that included millions of self-driving scenarios, classified according to the concepts appearing in each scene. One set included 500,000 scenarios, while the other had 3 million scenarios, with multiple concepts categorized for each scenario. Together, these scenarios provided hundreds of millions of data points to train the concept classifier, enabling the system to recognize diverse types of driving situations.
When evaluating the system's performance on test data, the accuracy of concept classification reached 54 percent, with a recall rate of 77 percent, and an accuracy of 23 percent. These results demonstrated the system's ability to link the decision-making process to concepts that are human-understandable, according to the team.
The researchers did not limit themselves to testing during simulation, but also deployed the “CW-Net” system on a real self-driving car belonging to the “Motional” company, during experiments conducted on a special track, and in the presence of a safety driver. During these experiments, the car faced surprising situations, as the safety drivers were asked to monitor its behavior and try to predict what it would do. The results showed that the explanations provided by the system improved drivers' ability to form a more accurate perception of how the car operates, and thus better predict its behavior, especially in unexpected situations.
A new driving style
Dr. Ewen Kenny, the lead researcher for the study at the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology, explains that the system differs from traditional methods used to interpret self-driving car decisions; It not only explains the decision after it is made, but rather links the interpretation directly to the decision-making process within the leadership system itself. The system transforms the complex thinking mechanism in the deep learning model into clear concepts, such as “approaching a parked vehicle,” or “approaching a cyclist,” and actually uses these concepts when determining the car’s path, making the interpretation causally linked to the decision, and not just an external description that may be misleading, with the possibility of presenting it in real time without affecting the performance of the driving system.
Kenny added to Asharq Al-Awsat that these concepts actually enter into the planning and decision-making process, which gives the interpretation a real causal link to the car’s behavior, far from being an external description that may be misleading. He pointed out that the system is able to provide these explanations in real time along with the path that the car chooses, without any impact on the performance of the original driving model.
Kenny pointed out that tests conducted on a real self-driving vehicle showed that providing safety drivers with CW-Net interpretations made them more able to predict the car's behavior, especially in sudden situations. He gave an example of an incident in which the car stopped when it was approaching a cyclist. The safety driver initially thought that the stop came because the system correctly detected the cyclist. However, the interpretation of the system revealed a different truth, which is that the driving model was not dealing with the cyclist correctly, and that the stop actually occurred as a result of the intervention of the emergency braking system after the car got too close to it.
Kenny explained that knowing the real reason behind the stop changed the driver's understanding of what was actually happening, and allowed him to take action earlier, such as reducing speed, or switching to manual driving. This information can also provide engineers with a better way to detect and address errors in artificial intelligence systems.
Kenny pointed out that the next step is to expand the scope of concepts and situations that the system can handle, and to test different methods for training and designing it, to improve its performance and interpretability, in addition to conducting broader tests, and in more diverse driving conditions, to ensure the accuracy and reliability of its interpretations, stressing the need for these interpretations not to give drivers false confidence, but rather to accurately reflect the real reasons for the car’s decisions, explaining that the ultimate goal is to ensure that “CW-Net” is able to help humans understand and predict the behavior of self-driving vehicles. Reliably and consistently in various driving conditions.
