AI for Science seminar with Sarah Alamdari, Microsoft Research.
Overview
- Date:Starts 11 September 2025, 15:00Ends 11 September 2025, 16:30
- Seats available:40
- Location:
- Language:English
Zoom password: ai4science
The on-site event will be followed by fika in the Analysen coffee area (fika from 16:00-16:30).
Abstract:
Engineered proteins play increasingly essential roles in applications spanning pharmaceuticals, molecular tools, synthetic biology, and more. Deep generative models present an opportunity to accelerate protein engineering for therapeutic and biological applications. Recent advances in protein language models, has facilitated more controllable and flexible approaches to protein design.
In this talk I will cover our recent work in developing these general-purpose protein language models and exploring their capabilities for functional protein design. First, we will discuss a general-purpose diffusion framework, EvoDiff, that combines evolutionary-scale data with the distinct conditioning capabilities of diffusion models for controllable protein design. Next, we introduce an extension of this work titled Dayhoff, a generative model of proteins that combines single sequences and sequence alignments into one autoregressive model via long context lengths.
In this work, we explore how model size, data quality, and diversity impact gains in protein expression. We envision that these modeling frameworks will enable new capabilities in protein engineering towards programmable, functional design.
About the speaker:
Sarah Alamdari is currently a Senior Applied Scientist at Microsoft Research New England, in the Biomedical Machine Learning Group. Here, she uses artificial intelligence to design new proteins, with the goal of advancing our understanding biology. Her recent work has focused on developing new protein language models, employing machine learning techniques on protein sequence datasets.
Previously, she completed her PhD in Chemical Engineering at the University of Washington as an NSF graduate research fellow where she applied and developed molecular dynamic frameworks to study the behavior of biomolecules at complex interfaces.
Structured learning
This theme focuses on how to make use of structure in data to build machine learning (ML) and artificial intelligence (AI) systems which are safer, more trustworthy and generalize better. Structure includes the relationship between data, in time and space, and how the predictions change when data is transformed in specific ways, for example rotated or scaled. These topics are abstract and general but have a direct impact on the use of AI and ML in the sciences and in applications such as drugs and materials design, or medical imaging.
