Download Research Tools
Researchers believe that pathogens are evolving to evade detection from the human immune system. I recently co-published a paper that discussed research into the ongoing evolutionary struggle between the immune system and pathogens. In this study, we sought to identify possible commonalities in HLA (human leukocyte antigen) binding preferences that would reveal patterns of optimization of this component of the immune system in response to the variation in pathogens.
I worked with post-doctoral student Tomer Hertz (now with Fred Hutchinson Cancer Research Lab) and a distinguished group of colleagues from the Institute for Immunology and Infectious Diseases, Royal Perth Hospital, and Murdoch University (Western Australia); the School of Anatomy and Human Biology, Centre for Forensic Science, University of Western Australia (Western Australia); and Fundacion Ciencia para la Vida (Chile).
Our paper, "Mapping the Landscape of Host-Pathogen Coevolution: HLA Class I Binding and Its Relationship with Evolutionary Conservation in Human and Viral Proteins," appeared in the American Society for Microbiology's Journal of Virology in February 2011. I'd like to share some highlights from the study with you.
Identifying Possible Commonalities in HLA Binding Preference
The majority of the cells in our bodies express something called HLA molecules, whose role is to sample cellular proteins and present them on the cellular surface for external surveillance by the specialized cells of our immune systems. This action forces all cells to reveal imprints of their inner workings.
When something out of the ordinary is detected—for example, the presence of an unusual mutation or a gene expression—the type and quality of the presented samples can spur the immune system's specialized killer cells into action. By sending kill signals to "odd" cells, the immune system can stop diseases such as cancer or viral infections. (Viruses bring their own genetic material to the cell and use the cellular resources to propagate.)
However, this scrutiny of the immune system creates evolutionary pressure on viruses, which often mutate to evade detection. Since the system of HLA molecules is highly selective in its sampling of protein segments, the mutational patterns in viruses are not entirely random: mutations tend to occur within the segments that HLA molecules are most likely to present.
On the other hand, over many generations, the distribution of thousands of HLA variants present in human populations may change. Additionally, in different geographic regions, we find significant variation in frequencies of different HLA molecules. This sets up an evolutionary game between the viruses on the one side and our immune systems on the other.
In order to analyze the results of the evolutionary processes that are driven by the interaction of HLA molecules with a wide diversity of viral intruders, we quantified the HLA binding preferences by using a novel measure called "targeting efficiency."
Targeting efficiency entails capturing the correlation between HLA-peptide binding affinities with the genetic conservation in the targeted proteomic regions. If HLA molecules possessed such targeting efficiency, this would (presumably) prove beneficial to humans. In theory, HLA molecules would draw attention to protein segments that are shared across related viral species as functionally important and thus immutable sections of their proteins. Individual invading viruses would find it more difficult to evade surveillance by mutating, because mutation within these segments would ruin the protein function. Targeting efficiency could even allow the immune system to generalize across related viral species.
Our analysis of targeting efficiencies for 95 HLA Class I alleles over thousands of human proteins and 52 human viruses indicate that HLA molecules do indeed prefer to target conserved regions in these proteomes! However, the arboviral Flaviviridae (for example, Dengue virus) proved a notable exception in which non-conserved regions were the preferred target of most alleles.
HLA molecules are encoded in three separate parts of the human genome: A, B, and C. During our study, we discovered that the oldest versions of our HLA molecules—namely the HLA-A alleles and several HLA-B alleles that had maintained a close sequence identity with chimpanzee homologues—were targeting conserved human proteins and DNA viruses (for example, Herpesviridae and Adenoviridae) most efficiently.
By contrast, the HLA-B alleles were targeting RNA viruses efficiently. This is reminiscent of predator-prey patterns that have been identified in evolutionary theory. For example, we know the following factors to be true:
Based on this information, we can extrapolate that evolution is going to drive their binding properties in different directions, thus splitting their targets, as in the established Lotka-Volterra (predator-prey) model of different types of foxes and rabbits inhabiting the same forest. In addition, we identified various patterns of host/pathogen specialization that are consistent with co-evolutionary selection and were also functionally relevant in specific cases. For example, preferential HLA targeting of conserved proteomic regions is associated with improved outcomes in HIV infections as well as protection against Dengue Hemorrhagic Fever.
I have just scratched the surface of the study in this blog. For complete study details, including a complete presentation of our methodology and findings, please follow the links below.
—Nebojsa Jojic, Principal Researcher, Microsoft Research eScience Group
Back in January, I blogged about Project Hawaii, a research and academic outreach program sponsored by Microsoft Research in cooperation with 20 universities worldwide. Approximately 300 students at those universities are developing applications for Windows Phone 7 this semester as part of the program. These students have already come up with new and innovative scenarios by using our previously released Relay and Rendezvous services. Beginning today, they will have another cloud service in their development arsenal: a Speech to Text Service.
This new cloud service will enable Project Hawaii participants to expand their applications with options such as diction, transcription, and voice commands. Students will also be able to use the new service to integrate other complex applications, such as Microsoft Translator, into their development projects. There is one limitation: Speech to Text currently supports English only. There are no plans to expand into other languages at this time.
In addition to making this service available to our Project Hawaii students, we are also releasing sample code from an application for Windows Phone 7 as part of the software development kit (SDK). This sample will allow users to speak into a phone and get transcribed text of their words in return. Plus, we'll be releasing an Optical Character Recognition (OCR) service for our Hawaii participants to use in the near future.
—Arjmand Samuel, Research Project Manager with the Microsoft Research Connections division of Microsoft Research
The neurosurgeon hovers over the patient, preparing to excise a life-threatening brain tumor. In this delicate operation, there is no margin for error: the tumor needs to be cut out with minimal damage to the surrounding healthy tissue. By using simple hand gestures, the surgeon signals a computer to display high-resolution scans of the patient’s brain, showing the physician where to place her scalpel, detailing the boundaries between diseased and healthy tissue. No longer must the neurosurgeon stop to refer to the patient’s image data during the operation, removing her gloves and potentially compromising the sterile surgical field. The upshot for the patient: reduced time under anesthesia and a lower risk of introduced infection.
Science fiction? Far from it. This scenario and others like it are on the verge of realization thanks to ground-breaking InnerEye project being conducted by Microsoft Research and a host of collaborators, including Johns Hopkins Medical Institute, The University of Oxford, Cornell Medical School, Massachusetts General Hospital, the University of Washington, Kings College London, and Cambridge University Hospitals.
The analysis of medical images is essential in modern medicine. As images have achieved higher and higher resolutions, the increasing amount of patient data has presented new challenges and opportunities, from diagnosis to therapy. The InnerEye research shows how a single, underlying image-recognition algorithm can enable a multitude of clinical applications, such as semantic image navigation, multimodal image registration, quality control, content-based image search, and natural user interfaces for surgery.
InnerEye takes advantage of advances in computer-human interactions that have put computers on a path to work for us and collaborate with us. The development of a natural user interface (NUI) enables computers to adapt to you and be more integrated into your environment via speech, touch, and gesture. As NUI systems become more powerful and are imbued with more situational awareness, they can provide beneficial, real-time interactions that will be seamless and naturally suited to your context—in short, systems will understand where you are and what you’re doing.
At this year’s TechFest—the annual event that showcases the latest work from Microsoft Research’s labs around the world—InnerEye is one of several projects that show where Microsoft is headed with NUI technologies, and how “futuristic” computing experiences are quickly becoming a reality. Building on the success of Kinect—a prime example of NUI technology reaching consumer scale—Microsoft Research continues to explore technologies that will enable the coming shift in how humans will communicate with machines, and vice versa. The possibilities are seemingly endless in how we approach the integration of computing into our lives and can enable a new era of creativity, social interaction, and technological scenarios.
—Antonio Criminisi, Researcher, Microsoft Research and Kristin Tolle, Director, Natural User Interface Team, Microsoft Research Connections division of Microsoft Research