BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.2.2.1//EN TZID:Asia/Jerusalem X-WR-TIMEZONE:Asia/Jerusalem BEGIN:VEVENT UID:1269@biology.technion.ac.il DTSTART;TZID=Asia/Jerusalem:20250303T130000 DTEND;TZID=Asia/Jerusalem:20250303T133000 DTSTAMP:20250226T142355Z URL:https://biology.technion.ac.il/en/seminars/m-sc-graduate-seminar-tal-i fargan-the-faculty-of-data-and-decision-sciences/ SUMMARY:M.Sc. Graduate Seminar-Tal Ifargan (The Faculty of Data and Decisio n Sciences) [No Categories] DESCRIPTION:Location: Hybrid- in the Faculty Auditorium/ZOOM:https://techni on.zoom.us/j/95712138406 Tal Ifargan\n Affiliation: \n Host:Prof. Kishony Roy \n AI-Driven Research and Trait Aware Representations of Medical Diag nostics\n\n \;\n\nAbstract\n\n \;\n\nThis seminar is devoted to tw o separate projects:\n\n \;\n\nAs AI promises to accelerate scientific discovery\, it remains unclear whether AI systems can perform fully auton omous research and whether they can do so while adhering to key scientific values\, such as transparency\, traceability and verifiability. Mimicking human scientific practices\, we built data-to-paper\, an automation platf orm that guides interacting LLM agents through a complete stepwise researc h process\, from annotated data to comprehensive research papers\, while p rogrammatically back-tracing information flow\, resulting in “data-chain ed” manuscripts. Testing the platform on diverse datasets\, it produced autonomously correct papers in 80%-90% of runs for simple datasets and res earch goals\, yet human interventions became critical for more complex tas ks. Our work demonstrates a potential for AI-driven acceleration of scient ific discovery in data-driven research and beyond\, while setting through “data-chaining” a new standard for verifiability and traceability for the coming era of AI-driven science.\n\n \;\n\nElectronic health recor ds offer significant potential for uncovering trait-specific patterns and advancing personalized medicine. Various methods\, mainly borrowed from na tural language processing\, have been proposed to represent International Classification of Diseases (ICD) codes\, yet these approaches often yield representations that are difficult to quantify and primarily serve predict ive tasks. Here\, we present TAR—a method for Trait-Aware Representation s of ICD codes implemented as a modified skip-gram model and demonstrate i ts ability to explore sex-specific differences in medical diagnostics\, a topic often overlooked in prior literature. In collaboration with Maccabi Healthcare Services\, we trained TAR on 25 years of data from 1.3 million patients\, revealing sex-dependent dynamics of aging. In addition\, the le arned embeddings encode the concept of sex as a distinct direction. Overal l\, TAR presents sex-specific diagnostic differences\, and is readily exte ndable to other phenotypic traits\, offering a versatile tool for broader biomedical discovery. LOCATION:Hybrid- in the Faculty Auditorium/ZOOM:https://technion.zoom.us/j/ 95712138406 END:VEVENT BEGIN:VTIMEZONE TZID:Asia/Jerusalem X-LIC-LOCATION:Asia/Jerusalem BEGIN:STANDARD DTSTART:20241027T010000 TZOFFSETFROM:+0300 TZOFFSETTO:+0200 TZNAME:IST END:STANDARD END:VTIMEZONE END:VCALENDAR