BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.4.0.1//EN
TZID:America/New_York
X-WR-TIMEZONE:America/New_York
BEGIN:VEVENT
UID:852@fds.yale.edu
DTSTART;TZID=America/New_York:20250319T130000
DTEND;TZID=America/New_York:20250319T140000
DTSTAMP:20250916T142147Z
URL:https://fds.yale.edu/events/fds-special-seminar-peiran-jin-microsoft-r
 esearch-nature-language-model-deciphering-the-language-of-nature-for-scien
 tific-discovery/
SUMMARY:FDS Special Seminar: Peiran Jin (Microsoft Research)\, "Nature Lang
 uage Model: Deciphering the Language of Nature for Scientific Discovery"
DESCRIPTION:\nNature Language Model: Deciphering the Language of Nature for
  Scientific Discovery [https://arxiv.org/abs/2502.07527]\n\n\n\nAbstract: 
 Foundation models have revolutionized natural language processing and arti
 ficial intelligence\, significantly enhancing how machines comprehend and 
 generate human languages. Inspired by the success of these foundation mode
 ls\, researchers have developed foundation models for individual scientifi
 c domains\, including small molecules\, materials\, proteins\, DNA\, RNA a
 nd even cells. However\, these models are typically trained in isolation\,
  lacking the ability to integrate across different scientific domains. Rec
 ognizing that entities within these domains can all be represented as sequ
 ences\, which together form the "language of nature"\, we introduce Nature
  Language Model (NatureLM)\, a sequence-based science foundation model des
 igned for scientific discovery. Pre-trained with data from multiple scient
 ific domains\, NatureLM offers a unified\, versatile model that enables va
 rious applications including: (i) generating and optimizing small molecule
 s\, proteins\, RNA\, and materials using text instructions\; (ii) cross-do
 main generation/design\, such as protein-to-molecule and protein-to-RNA ge
 neration\; and (iii) top performance across different domains\, matching o
 r surpassing state-of-the-art specialist models. NatureLM offers a promisi
 ng generalist approach for various scientific tasks\, including drug disco
 very (hit generation/optimization\, ADMET optimization\, synthesis)\, nove
 l material design\, and the development of therapeutic proteins or nucleot
 ides. We have developed NatureLM models in different sizes (1 billion\, 8 
 billion\, and 46.7 billion parameters) and observed a clear improvement in
  performance as the model size increases.\n\n\n\nSpeaker bio: Peiran Jin i
 s a Senior Research Engineer at Microsoft Research AI for Science team\, w
 here he focuses on LLM\, Diffusion model\, protein sequence and structure 
 modeling. His work explores the intersection of deep learning and molecula
 r modeling to protein structure prediction and related applications. Befor
 e joining Microsoft\, Peiran was a Senior Research Engineer at Seagate R&a
 mp\;D team\, where he contributed to the development of heat-assisted magn
 etic recording technology. He holds a PhD in Physics from Georgetown Unive
 rsity and a Bachelor’s degree from the University of Science and Technol
 ogy of China.\n\n\n\nHosted by Mark Gerstein and the Gerstein Lab.\n
CATEGORIES:FDS Events,Special Seminar
LOCATION:Bass Center for Molecular and Structural Biology\, Room 405\, 266 
 Whitney Avenue\, New Haven\, CT\, 06511\, United States
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=266 Whitney Avenue\, New Ha
 ven\, CT\, 06511\, United States;X-APPLE-RADIUS=100;X-TITLE=Bass Center fo
 r Molecular and Structural Biology\, Room 405:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
DTSTART:20250309T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR