AI
-
New Insights into How the Human Brain Solves Complex Decision-Making Problems
A new study on meta reinforcement learning algorithms helps us understand how the human brain learns to adapt to complexity and uncertainty when learning and making decisions. A research team, led by Professor Sang Wan Lee at KAIST jointly with John O’Doherty at Caltech, succeeded in discovering both a computational and neural mechanism for human meta reinforcement learning, opening up the possibility of porting key elements of human intelligence into artificial intelligence algorithms. This study provides a glimpse into how it might ultimately use computational models to reverse engineer human reinforcement learning.
This work was published on Dec 16, 2019 in the journal Nature Communications. The title of the paper is “Task complexity interacts with state-space uncertainty in the arbitration between model-based and model-free learning.”
Human reinforcement learning is an inherently complex and dynamic process, involving goal setting, strategy choice, action selection, strategy modification, cognitive resource allocation etc. This a very challenging problem for humans to solve owing to the rapidly changing and multifaced environment in which humans have to operate. To make matters worse, humans often need to often rapidly make important decisions even before getting the opportunity to collect a lot of information, unlike the case when using deep learning methods to model learning and decision-making in artificial intelligence applications.
In order to solve this problem, the research team used a technique called 'reinforcement learning theory-based experiment design' to optimize the three variables of the two-stage Markov decision task - goal, task complexity, and task uncertainty. This experimental design technique allowed the team not only to control confounding factors, but also to create a situation similar to that which occurs in actual human problem solving.
Secondly, the team used a technique called ‘model-based neuroimaging analysis.’ Based on the acquired behavior and fMRI data, more than 100 different types of meta reinforcement learning algorithms were pitted against each other to find a computational model that can explain both behavioral and neural data. Thirdly, for the sake of a more rigorous verification, the team applied an analytical method called ‘parameter recovery analysis,’ which involves high-precision behavioral profiling of both human subjects and computational models.
In this way, the team was able to accurately identify a computational model of meta reinforcement learning, ensuring not only that the model’s apparent behavior is similar to that of humans, but also that the model solves the problem in the same way as humans do.
The team found that people tended to increase planning-based reinforcement learning (called model-based control), in response to increasing task complexity. However, they resorted to a simpler, more resource efficient strategy called model-free control, when both uncertainty and task complexity were high. This suggests that both the task uncertainty and the task complexity interact during the meta control of reinforcement learning. Computational fMRI analyses revealed that task complexity interacts with neural representations of the reliability of the learning strategies in the inferior prefrontal cortex.
These findings significantly advance understanding of the nature of the computations being implemented in the inferior prefrontal cortex during meta reinforcement learning as well as providing insight into the more general question of how the brain resolves uncertainty and complexity in a dynamically changing environment. Identifying the key computational variables that drive prefrontal meta reinforcement learning, can also inform understanding of how this process might be vulnerable to break down in certain psychiatric disorders such as depression and OCD. Furthermore, gaining a computational understanding of how this process can sometimes lead to increased model-free control, can provide insights into how under some situations task performance might break down under conditions of high cognitive load.
Professor Lee said, “This study will be of enormous interest to researchers in both the artificial intelligence and human/computer interaction fields since this holds significant potential for applying core insights gleaned into how human intelligence works with AI algorithms.”
This work was funded by the National Institute on Drug Abuse, the National Research Foundation of Korea, the Ministry of Science and ICT, Samsung Research Funding Center of Samsung Electronics.
Figure 1 (modified from the figures of the original paper doi:10.1038/s41467-019-13632-1). Computations implemented in the inferior prefrontal cortex during meta reinforcement learning. (A) Computational model of human prefrontal meta reinforcement learning (left) and the brain areas whose neural activity patterns are explained by the latent variables of the model. (B) Examples of behavioral profiles. Shown on the left is choice bias for different goal types and on the right is choice optimality for task complexity and uncertainty. (C) Parameter recoverability analysis. Compared are the effect of task uncertainty (left) and task complexity (right) on choice optimality.
-Profile
Professor Sang Wan Lee
sangwan@kaist.ac.kr
Department of Bio and Brain Engineering
Director, KAIST Center for Neuroscience-inspired AI
KAIST Institute for Artificial Intelligence (http://aibrain.kaist.ac.kr)
KAIST Institute for Health, Science, and Technology
KAIST (https://www.kaist.ac.kr)
2020.01.31 View 5238 -
Professor Yim Decorated with the Chongjo Order of Merit
Professor Yong-Taek Yim from the Department of Mechanical Engineering was awarded the highest order of merit, the “Chongjo Keunjong Medal,” bestowed to public officials by the government in celebration of Invention Day on May 27.
Professor Yim was recognized for his innovative achievements to increase royalty income by introducing an IP-based management system at the Korean Institute of Machinery & Materials. He served as the president of KIMM for three years from 2014.
His idea led to new approaches to help explore diverse revenue creating sources such as dividend earnings and share sales, apart from simply relying on technology transfer fees. His efforts to disseminate the in-house R&D results also led to the foundation of six tech-based startups and spinoffs, which generated 11.2 billion KRW in sales. He also helped set up three spinoffs abroad.
Professor Yim said, “I pushed employee invention as a new value creator at KIMM. I thank each and every researcher and staff member at KIMM who worked so hard to create such an innovative IP-based R&D environment.”
2019.05.28 View 5774 -
KAIST Unveils the Hidden Control Architecture of Brain Networks
(Professor Kwang-Hyun Cho and his team)
A KAIST research team identified the intrinsic control architecture of brain networks. The control properties will contribute to providing a fundamental basis for the exogenous control of brain networks and, therefore, has broad implications in cognitive and clinical neuroscience.
Although efficiency and robustness are often regarded as having a trade-off relationship, the human brain usually exhibits both attributes when it performs complex cognitive functions. Such optimality must be rooted in a specific coordinated control of interconnected brain regions, but the understanding of the intrinsic control architecture of brain networks is lacking.
Professor Kwang-Hyun Cho from the Department of Bio and Brain Engineering and his team investigated the intrinsic control architecture of brain networks. They employed an interdisciplinary approach that spans connectomics, neuroscience, control engineering, network science, and systems biology to examine the structural brain networks of various species and compared them with the control architecture of other biological networks, as well as man-made ones, such as social, infrastructural and technological networks.
In particular, the team reconstructed the structural brain networks of 100 healthy human adults by performing brain parcellation and tractography with structural and diffusion imaging data obtained from the Human Connectome Project database of the US National Institutes of Health.
The team developed a framework for analyzing the control architecture of brain networks based on the minimum dominating set (MDSet), which refers to a minimal subset of nodes (MD-nodes) that control the remaining nodes with a one-step direct interaction. MD-nodes play a crucial role in various complex networks including biomolecular networks, but they have not been investigated in brain networks.
By exploring and comparing the structural principles underlying the composition of MDSets of various complex networks, the team delineated their distinct control architectures. Interestingly, the team found that the proportion of MDSets in brain networks is remarkably small compared to those of other complex networks. This finding implies that brain networks may have been optimized for minimizing the cost required for controlling networks. Furthermore, the team found that the MDSets of brain networks are not solely determined by the degree of nodes, but rather strategically placed to form a particular control architecture.
Consequently, the team revealed the hidden control architecture of brain networks, namely, the distributed and overlapping control architecture that is distinct from other complex networks. The team found that such a particular control architecture brings about robustness against targeted attacks (i.e., preferential attacks on high-degree nodes) which might be a fundamental basis of robust brain functions against preferential damage of high-degree nodes (i.e., brain regions).
Moreover, the team found that the particular control architecture of brain networks also enables high efficiency in switching from one network state, defined by a set of node activities, to another – a capability that is crucial for traversing diverse cognitive states.
Professor Cho said, “This study is the first attempt to make a quantitative comparison between brain networks and other real-world complex networks. Understanding of intrinsic control architecture underlying brain networks may enable the development of optimal interventions for therapeutic purposes or cognitive enhancement.”
This research, led by Byeongwook Lee, Uiryong Kang and Hongjun Chang, was published in iScience (10.1016/j.isci.2019.02.017) on March 29, 2019.
Figure 1. Schematic of identification of control architecture of brain networks.
Figure 2. Identified control architectures of brain networks and other real-world complex networks.
2019.04.23 View 37454 -
Wafer-Scale Multilayer Fabrication of Silk Fibroin-Based Microelectronics
A KAIST research team developed a novel fabrication method for the multilayer processing of silk-based microelectronics. This technology for creating a biodegradable silk fibroin film allows microfabrication with polymer or metal structures manufactured from photolithography. It can be a key technology in the implementation of silk fibroin-based biodegradable electronic devices or localized drug delivery through silk fibroin patterns.
Silk fibroins are biocompatible, biodegradable, transparent, and flexible, which makes them excellent candidates for implantable biomedical devices, and they have also been used as biodegradable films and functional microstructures in biomedical applications. However, conventional microfabrication processes require strong etching solutions and solvents to modify the structure of silk fibroins.
To prevent the silk fibroin from being damaged during the process, Professor Hyunjoo J. Lee from the School of Electrical Engineering and her team came up with a novel process, named aluminum hard mask on silk fibroin (AMoS), which is capable of micropatterning multiple layers composed of both fibroin and inorganic materials, such as metal and dielectrics with high-precision microscale alignment. The AMoS process can make silk fibroin patterns on devices, or make patterns on silk fibroin thin films with other materials by using photolithography, which is a core technology in the current microfabrication process.
The team successfully cultured primary neurons on the processed silk fibroin micro-patterns, and confirmed that silk fibroin has excellent biocompatibility before and after the fabrication process and that it also can be applied to implanted biological devices.
Through this technology, the team realized the multilayer micropatterning of fibroin films on a silk fibroin substrate and fabricated a biodegradable microelectric circuit consisting of resistors and silk fibroin dielectric capacitors in a silicon wafer with large areas.
They also used this technology to position the micro-pattern of the silk fibroin thin film closer to the flexible polymer-based brain electrode, and confirmed the dye molecules mounted on the silk fibroin were transferred successfully from the micropatterns.
Professor Lee said, “This technology facilitates wafer-scale, large-area processing of sensitive materials. We expect it to be applied to a wide range of biomedical devices in the future. Using the silk fibroin with micro-patterned brain electrodes can open up many new possibilities in research on brain circuits by mounting drugs that restrict or promote brain cell activities.”
This research, in collaboration with Dr. Nakwon Choi from KIST and led by PhD candidate Geon Kook, was published in ACS AMI (10.1021/acsami.8b13170) on January 16, 2019.
Figure 1. The cover page of ACS AMI
Figure 2. Fibroin microstructures and metal patterns on a fibroin produced by using the AMoS mask.
Figure 3. Biocompatibility assessment of the AMoS Process. Top: Schematics image of a) fibroin-coated silicon b) fibroin-pattered silicon and c) gold-patterned fibroin. Bottom: Representative confocal microscopy images of live (green) and dead (red) primary cortical neurons cultured on the substrates.
2019.03.15 View 21280 -
1g-Ultrasound System for the Brain Stimulation of a Freely-moving Mouse
A KAIST research team developed a light-weight capacitive micromachined ultrasonic transducer (CMUT) and succeeded in the ultrasound brain stimulation of a freely-moving mouse. With this lightweight and compact system, researchers can conduct a versatile set of in vivo experiments.
Conventional methods for stimulating a specific brain region, such as deep brain stimulation (DBS) and optogenetics technology, are highly invasive because they have to insert probes into a target brain, which makes them difficult to use for clinical application. While transcranial magnetic stimulation (TMS) and transcranial electrical stimulation (TES) are noninvasive, they have a wide range of stimulation and problems with in-depth stimulation, which makes them problematic for target-specific treatment.
Therefore, noninvasive and focused ultrasound stimulation technology is gaining a great deal of attention as a next-generation brain stimulation alternative. Since it is delivered noninvasively, it can be applied safely in humans as well as animal experiments. Focused ultrasound stimulation is more advantageous than conventional methods in terms of providing both local and deep stimulation.
Animal behavior experiments are essential for brain stimulation research; however, ultrasonic brain stimulation technology is currently in the early stages of development. So far, only research outcomes with fixed anesthetized mice have been studied because of the heavy ultrasonic device.
Professor Hyunjoo J. Lee from the School of Electrical Engineering and her team reported a technology that can provide ultrasound stimulation to the brain of a freely-moving mouse through a microminiaturized ultrasound device.
The team studied miniaturization and ultra-lightweight CMUTs through microelectromechanical systems (MEMS) technology and designed a device suitable for behavior experiments. The device weighing less than 1g (around 0.05% of the mouse’s weight) has the center frequency, size, focal length, and ultrasonic intensity to fit a mouse’s dimensions.
To evaluate the performance of the ultrasonic device, the team stimulated the motor cortex of the mouse brain and observed the movement reaction of its forefoot. They also measured the electromyography (EMG) of the trapezius.
As a result, the team confirmed that their ultrasonic device can deliver ultrasound to a depth of 3-4mm in the mouse brain and stimulate an area of the mouse brain that represents 25% of its total size.
Based on this research, the team is investigating the effects of ultrasound on sleep by stimulating the brain of sleeping mice.
Professor Lee said, “Going beyond experimenting on fixed anesthetized mice, this research succeeded in the brain stimulation of a freely-moving mouse. We are planning to study mice with diseases, such as Parkinson’s disease, dementia, depression, and epilepsy. I believe that this basic research can contribute to treating human brain-related diseases through ultrasound brain stimulation.
This research, led by Masters candidates Hyunggug Kim and Seongyeon Kim, was published in Brain Stimulation (10.1016/j.brs.2018.11.007) on November 17, 2018.
Figure 1. The miniature transducer for the transcranial ultrasound of a freely-moving mouse
Figure 2. Its structure and simulated 2D beam profile in the axial ad radial directions
2019.03.13 View 8594 -
Noninvasive Light-Sensitive Recombinase for Deep Brain Genetic Manipulation
A KAIST team presented a noninvasive light-sensitive photoactivatable recombinase suitable for genetic manipulation in vivo. The highly light-sensitive property of photoactivatable Flp recombinase will be ideal for controlling genetic manipulation in deep mouse brain regions by illumination with a noninvasive light-emitting diode. This easy-to-use optogenetic module made by Professor Won Do Heo and his team will provide a side-effect free and expandable genetic manipulation tool for neuroscience research.
Spatiotemporal control of gene expression has been acclaimed as a valuable strategy for identifying functions of genes with complex neural circuits. Studies of complex brain functions require highly sophisticated and robust technologies that enable specific labeling and rapid genetic modification in live animals. A number of approaches for controlling the activity of proteins or expression of genes in a spatiotemporal manner using light, small molecules, hormones, and peptides have been developed for manipulating intact circuits or functions.
Among them, recombination-employing, chemically inducible systems are the most commonly used in vivo gene-modification systems. Other approaches include selective or conditional Cre-activation systems within subsets of green fluorescent protein-expressing cells or dual-promoter-driven intersectional populations of cells.
However, these methods are limited by the considerable time and effort required to establish knock-in mouse lines and by constraints on spatiotemporal control, which relies on a limited set of available genetic promoters and transgenic mouse resources.
Beyond these constraints, optogenetic approaches allow the activity of genetically defined neurons in the mouse brain to be controlled with high spatiotemporal resolution. However, an optogenetic module for gene-manipulation capable of revealing the spatiotemporal functions of specific target genes in the mouse brain has remained a challenge.
In the study published at Nature Communication on Jan. 18, the team featured photoactivatable Flp recombinase by searching out split sites of Flp recombinase that were not previously identified, being capable of reconstitution to be active. The team validated the highly light-sensitive, efficient performance of photoactivatable Flp recombinase through precise light targeting by showing transgene expression within anatomically confined mouse brain regions.
The concept of local genetic labeling presented here suggests a new approach for genetically identifying subpopulations of cells defined by the spatial and temporal characteristics of light delivery. To date, an optogenetic module for gene-manipulation capable of revealing spatiotemporal functions of specific target genes in the mouse brain has remained out of reach and no such light-inducible Flp system has been developed. Accordingly, the team sought to develop a photoactivatable Flp recombinase that takes full advantage of the high spatiotemporal control offered by light stimulation.
This activation through noninvasive light illumination deep inside the brain is advantageous in that it avoids chemical or optic fiber implantation-mediated side effects, such as off-target cytotoxicity or physical lesions that might influence animal physiology or behaviors. The technique provides expandable utilities for transgene expression systems upon Flp recombinase activity in vivo, by designing a viral vector for minimal leaky expression influenced by viral nascent promoters.
The team demonstrated the utility of PA-Flp as a noninvasive in vivo optogenetic manipulation tool for use in the mouse brain, even applicable for deep brain structures as it can reach the hippocampus or medial septum using external LED light illumination.
The study is the result of five years of research by Professor Heo, who has led the bio-imaging and optogenetics fields by developing his own bio-imaging and optogenetics technologies. “It will be a great advantage to control specific gene expression desired by LEDs with little physical and chemical stimulation that can affect the physiological phenomenon in living animals,” he explained.
2019.01.22 View 6112 -
KAIST Presents Innovations at CES 2019
Ten of the most innovative technologies spun off from KAIST made a debut at the Consumer Electronics Show (CES) 2019, the world’s largest consumer electronics and IT exhibition being held in Las Vegas from January 8 to 11. The KAIST booth at the CES featured technologies made by KAIST research teams and five startup companies including LiBEST, Memslux, and Green Power. In particular, the KAIST Alumni Association invited 33 aspiring alumni entrepreneurs selected from the KAIST Startup Competition to the show.
At the exhibition, KAIST is presenting innovations in the fields of AI and Bio-IT convergence for the Fourth Industrial Revolution. These include real-time upscaling from Full HD to 4K UHD using AI deep learning-based convolutional neural networks (Professor Munchurl Kim, School of Electrical Engineering) and an AI conversation agent that responds to user’s emotions (Professor Soo-Young Lee, School of Electrical Engineering).
Other technologies include optimal drug target identification by cancer cell type through drug response prediction to be used in personalized cancer treatments (Professor Kwang-Hyun Cho, Department of Bio and Brain Engineering), a nanofiber-based color changing gas sensor with greater sensitivity than conventional paper-based color changing sensors (Professor Il-Doo Kim, Department of Materials Science and Engineering), and functional near-infrared spectroscopy (fNIRS) for brain imaging and muscle fatigue measurement (Professor Hyeonmin Bae, School of Electrical Engineering).
The KAIST booth also features startups founded by KAIST alumni including LiBEST with a flexible lithium polymer secondary cell optimized for smart wearable devices and Rempus with a high-performance lithium ion cell packaging technology for outstanding safety, high capacity, long life, and fast charging.
Green Power and Smart Radar Systems are also joining the booth with a highly efficient and eco-friendly wireless charging system for electrical cars, and a 4D image radar sensor that detects 3D images and speed in real time for applications in self-driving cars, drones, and security systems respectively.
Faculty-founded startup Memslux (CEO Jun-Bo Yoon, School of Electrical Engineering) is presenting a transparent surface light source solution for next-generation display devices.
Associate Vice President of Office of University-Industry Cooperation Kyung Cheol Choi said, “I believe that universities should play a role in connecting technological innovations to business startups for creating value at a global level. In that sense, it is a great opportunity to present innovative technologies from KAIST and promote outstanding KAIST startups at CES 2019. Hopefully, this experience will lead to joint R&D, investment, cooperation, and international technology transfer contracts with leading companies from around the world.”
Here are the five key technologies presented by KAIST at CES 2019.
2019.01.10 View 7022 -
Vietnamese Alumni of Korean S&T Universities Gather in Hanoi
(Vietnamese KAIST alumni gather in Hanoi on November 24.)
(Dr.Huong Minh Nguyen at the Institute of Biotechnology in the Vietnam Academy of Science and Technology (VAST). She is the representative of Vietnamese KAIST alumni.)
Some came from Ho Chi Minh and some even flew in from Singapore. For those KAIST alumni who gathered in Hanoi, the trip was well worth it. More than 100 Vietnamese alumni of KAIST, GIST, and UST attended the reunion in Hanoi on November 24. Presidents and vice president from three universities welcomed them and celebrated their successful careers after returning home or starting careers in other countries.
The reunion was co-hosted by KAIST, UST, and GIST in an effort to make a platform for continued networking for scientists who have studied at Korea’s science and technology universities. This joint reunion will be expected to include other science and technology universities and institutes in the future.
Among 1,873 international KAIST alumni from 106 countries, the number of Vietnamese graduates is the most dominant with 262 alumni, 14% of the total international alumni. Welcoming them, KAIST Vice President Soohyun Kim said that he was very impressed that all of the alumni are making a very impressive stride in their fields.
“You will be a big asset to make your country grow. You will also be a bridge for future collaborations with your institutions and KAIST and Korea. Vietnam holds great potential for future prosperity especially in science and technology and we look forward to seeing this network continue to benefit both countries.” Vice President Kim said all of the presidents shared the idea to make this gathering a regular event. “Other S&T universities will join to hold joint reunions in other countries in the future,” he added.
Dr. Huong Minh Nguyen at the Institute of Biotechnology in the Vietnam Academy of Science and Technology (VAST) is one of 22 KAIST alumni who joined the reunion and supported the idea. She is also the representative of Vietnamese KAIST alumni. Dr. Nguyen, who earned her MS and PhD in the Department of Biological Sciences in 2013, spent five and half years at KAIST. “All the members at the lab were taking care of each other. When I joined the lab back then, they treated me as a ‘baby sister’ in a family and our relationship grew like a sister and brotherhood. I still appreciate the bonding relationship we could make with our colleagues.”
She cites a supporting culture, a competent educational system that places a focus on practicality, and rich resources that help researchers try whatever they want as the distinct advantages of a KAIST education.
“I never heard my professor saying “No” to any of my suggestions in conducting research. Professors and lab members put all their efforts, resources, and facilities to getting my research started. They helped me obtain all the resources that I needed. It was a huge encouragement to me,” she added.
She started her Masters, along with four Vietnamese colleagues, right after graduating from the Hanoi University of Natural Sciences, a top science university in Vietnam. Their experience at KAIST during a 4-week exchange program during undergraduates made them decide their academic destination without any doubt. All of them finished their PhDs in 2013. Two of them moved back home and the two others are now in the US and Germany as postdoctoral fellows.
“At that time, I could go to other countries for my further studies. But I already experienced KAIST for a month when I was an undergraduate, so I was not hesitant to go to KAIST. All of the classes are in English so, for students who do not speak in Korean, language does not bring any problems in studying and conducting research.”
She said that KAIST alumni are enjoying very successful careers in Vietnam and many foreign countries. “We do not have any problems choosing our careers back home and other countries we wanted to work in.”
However, like many PhD candidates who feel pressure about their studies and an uncertain future, her days at KAIST also included challenging times while adjusting to Korean culture. She took Korean classes for three semesters voluntarily at the KAIST Language Center to better understand Korea and her lab members. Her efforts paid off well. She could easily communicate with her colleagues and felt she became a real part of the inner group. But the stress remained to prove herself in research and she still had to deal with some bias.
“I know some people think Vietnam is behind Korea. Many people think that we are not as good as Korean students because we are from Vietnam. I desperately wanted to prove that I am as good as my peers.” Studying together with Korean and other international PhD candidates, she realized that everyone has their own purposes and pressures. “Even though there are minor differences, every PhD candidate has the same issues with their uncertain futures. It was quite comforting when we shared that it’s not only my problem. To understand that it’s a problem we all share comforts us a lot and we came to support each other.
To better help students release such pressure and stress, she said the university needs to create more diverse institutional channels to communicate with them. “Looking back, I was younger and less competent to speak up about when we were stressed and needed to ask for help. I hope students can begin to release their pressure and stress through diverse channels and resolve the problems,” she said.
Asked about her future plans, she replied, “I don’t think I can do anything better than what I am doing now. I enjoy what I am doing now at my institute. But in the near very future, I want to visit Korea and KAIST again.”
2018.11.27 View 3299 -
A Breakthrough for Understanding Glioblastoma: Origin Cells for Deadly Brain Tumors Identified
Figure 1. The pattern of GBM genesis is similar to that of firework. The bottom canon represents the first occurrence of the SVZ mutated cell.
A new study by KAIST researchers identified where the mutation causing glioblastoma starts. According to the study, neural stem cells away from the tumor mass are the cells of origin that contain mutation drivers for glioblastoma, one of the most aggressive brain tumor. This breakthrough research, reported in Nature on August 1, gives insights for understanding why glioblastomas almost always grow back, even after surgery, and suggests novel ways to treat glioblastoma, which was previously thought to be incurable.
Like most cancers, glioblastoma is treated with surgery to remove as much of the tumor as possible, then radiation and chemotherapy. However, it almost always returns in less than a year and its median survival time is only 15 months. Precision therapeutic approaches targeting tumors themselves didn’t lead to any breakthroughs.
Professor Jeong Ho Lee’s team at the Graduate School of Medical Science and Engineering described direct genetic evidence through the deep sequencing of all triple-matched samples: normal SVZ tissue away from the tumor mass, tumor tissue, and normal cortical tissue. The research team studied 28 patients with glioblastomas and other types of brain tumors who underwent supra-total resection or other surgical resections of tumors, providing access to normal subventricular zone (SVZ) tissue (where neural stem cells are located) away from the tumor mass. The researchers used various deep and single cell sequencing technologies to conduct comparative DNA analysis on the samples from the patient’s SVZ tissue and tumors.
They reported that normal SVZ tissue away from the tumor in 56.3% of patients with glioblastoma already contained low-level glioblastoma driver mutations that were observed at high levels in their matching tumors. Furthermore, the research team generated a genome edited mouse carrying glioblastoma mutations in the SVZ and showed that neural stem cells with mutations migrate from the SVZ lead to the development of glioblastomas in distant brain regions. (See the image below)
Professor Lee conducted this study in collaboration with Professor Seok-Gu Kang of the Brain Tumor Center at Severance Hospital of Yonsei University. He said, “It’s easier to understand when we compare it to fireworks. Every flare flying around sky can be likened to cancer cells even though the fireworks are triggered on the ground. We found the trigger.” The identification of this mutation pathway of glioblastomas will lead to a new paradigm for therapeutic strategies. He added, “Now, we can focus on interrupting the recurrence and evolution of glioblastomas.”
Professor Lee has investigated mutations arising in the brain for a decade. He is developing innovative diagnostics and therapeutics for untreatable brain disorders including intractable epilepsy and glioblastoma at a tech-startup, SoVarGen. “All technologies we used during the research were transferred to the company. This research gave us very good momentum to reach the next phase of our startup,” he remarked.
Figure 2. Genetic analysis of tumor-free SVZ tissue and matching tumor tissue from GBM patients.
Figure 3. Glioma progression in genome edited mice carrying GBM mutations in the SVZ
2018.08.02 View 11494 -
KAIST Partners with Taiwan for PhD Scholarship Program
(President Shin, Taiwanese Acting Minister of Education Yao, Deputy Director General Chang at the Ministry of Education pose after signing the MOU.(from left))
President Sung-Chul Shin signed an MOU with the Ministry of Education in Taiwan for the Taiwanese PhD scholarship program. The signing was made between President Shin and Dr. Yao Leehter, acting Minister of Education in Taiwan, on June 27.
The Taiwanese Ministry of Education is signing MOUs with prestigious universities around the world to encourage its students to pursue study abroad opportunities at top schools. According to the MOU, Taiwanese PhD candidates will be able to use KAIST scholarships for their tuition fees, while the Taiwanese government will provide a stipend and living costs for four years from next September.
KAIST became the 14th university to sign this MOU, joining a group of top universities that includes University of Cambridge, Oxford University, California Institute of Technology, and Columbia University in the US. KAIST is the first institution in Asia to sign the MOU.
Acting Minister Yao said, “KAIST has emerged as a world leading university in less than five decades since its foundation. This remarkable growth led us to partner with KAIST. We hope this will serve as an opportunity to further our partnership in research collaboration as well as students exchanges.”
President Shin appreciated the Taiwanese government’s recognition of KAIST’s global reputation. He said, “We will closely collaborate with the Taiwan government and its universities for transforming educational opportunities to better respond to the Fourth Industrial Revolution.
2018.06.27 View 5060 -
Czech Technology Mission with KAIST
Members of the Czech research community visited KAIST to discuss medium to long-term cooperation with KAIST. This visit was hosted by the Fourth Industrial Revolution Intelligence Center (FIRIC).
The community is comprised of people from Czech enterprises and academic institutes that are leading core technologies for the Fourth Industrial Revolution in the fields of AI, robotics, and biotechnology. They had a chance to meet KAIST professors and visit research labs.
Professor Il-Doo Kim from the Department of Materials Science and Engineering, Professor Seongsu Kim from the Department of Mechanical Engineering, and Professor Hyun Uk Kim from the Department of Chemical and Biomolecular Engineering attended the meeting, which took place in the Mechatronics, Systems, and Control Lab under the Vice President for Planning and Budget Soo Hyun Kim and Professor Kyung Soo Kim from the Department of Mechanical Engineering.
Professor Petr Novák from the Technical University of Ostrava said, “It was a meaningful meeting to help understand research trends on industrial robots in Korea.” Professor So Young Kim from FIRIC said, “The Czech research community is strong in basic research where KAIST has outstanding source technology. I hope this visit will open up a path for medium to long-term cooperation on sharing research and technology know-how between the Czech research community and KAIST.”
2018.06.07 View 5602 -
Fast-Charging Lithium-Oxygen Batteries
(Professor Hye Ryung Byon)
KAIST researchers have paved the way for fast-charging lithium-oxygen batteries.
Professor Hye Ryung Byon from the Department of Chemistry and Professor Yousung Jung from the Graduate School of EEWS led a joint research team to develop lithium-oxygen batteries exhibiting 80% round-trip efficiency even at high charging rates, solving the problem of existing lithium-oxygen batteries which generally showed drastically lower efficiencies when the charge current rate was increased.
This study exploits the size and shape lithium peroxide, a discharge product, which is known to cause the very problems mentioned above. In doing so, the researchers have lowered the overpotential, which is the difference between the thermodynamic reversible potential and the measured potential, and simultaneously improved battery efficiency. Of particular interest is the fact that these high-performance lithium-oxygen batteries can be realized without costly catalysts.
One remarkable property of lithium-oxygen batteries is that they can accommodate three to five times the energy density of lithium-ion batteries commonly used today. Therefore, lithium-oxygen batteries would render longer driving distance to electric vehicles or drones, which operate on the continued use of electrical power.
However, their weakness lies in that, during charge, the lithium peroxide remains undecomposed at low overpotential, resulting in eventually compromising the battery’s overall performance. This is due to the poor ionic and electrical conductivity of lithium peroxide.
To tackle this issue, the researchers could form one-dimensional amorphous lithium peroxide nanostructures through the use of a mesoporous carbon electrode, CMK-3. When compared against non-mesoporous electrodes, CMK-3 showed exceptionally lower overpotential, thereby enhancing the round-trip efficiency of lithium-oxygen batteries.
The amorphous lithium peroxide produced along the electrode has a small volume and a large surface area contacting electrolyte solution, which is presumably endowed with high conductivity to speed up the charging of the lithium-oxygen batteries.
This research underpins the feasibility of overcoming the fundamental limitations of lithium-oxygen batteries even without the addition of expensive catalytic materials, but rather by the re-configuration of the size and shape of the lithium peroxide.
The findings of this research were published in Nature Communications on February 14.
Figure 1. Transmission electron microscopy (TEM) images
Figure 2. Galvanostatic rate capability
Figure 3. Density functional calculation and Bader charge analysis
2018.05.30 View 8855