Integrated Calendar

Today, enzymes can be reprogrammed beyond biology's natural reactions. This is achieved through noncovalent interactions decorating and templating enzyme active sites to stabilize a single transition state amid competing alternatives. We engineered a flavin-dependent oxidoreductase to generate tertiary radicals within its active site and channel them into stereoselective C–N coupling with unsubstituted anilines, delivering chiral α-tertiary amines with good yields and high chemo- and enantioselectivity, under visible light irradiation—without metal cofactors.Six rounds of directed evolution install a π-stacking/hydrogen-bond network that templates lone pair–radical hyperconjugation, overriding arene addition to enforce C(sp³)–N formation. DFT and multivariate statistical modeling reveal how this microenvironment flips innate radical reactivity toward N-alkylation. This establishes biocatalytic spatial programming as a general platform for non-natural radical transformations beyond both evolution and small-molecule catalysis.

First-passage processes are pervasive in the world around us, yet a general framework for understanding their response to external perturbations remains elusive. In this talk, I will present a universal linear response theory for first-passage times, showing that the response of the mean first-passage time to a rare perturbation depends only on the unperturbed mean and variance, and the mean completion time after perturbation onset. I will demonstrate how this simple, yet powerful, result offers a tool for predicting the impact of perturbations on first-passage processes across various scenarios, including non-trivial ones such neural network training. I will also demonstrate how the result can be used in reverse to infer fluctuations from mean first-passage times, opening the door for the extraction of single-molecule fluctuations from bulk (concentration) measurements.FOR THE LATEST UPDATES AND CONTENT ON SOFT MATTER AND BIOLOGICAL PHYSICS AT THE WEIZMANN, VISIT OUR WEBSITE: https://www.bio

Modern text-to-image models achieve strong visual fidelity and prompt alignment, but often at the cost of generative diversity.

In this talk, I will present two complementary approaches to this problem: a simple inference-time method that achieves rich diversity by intervening in the model’s internal representations, and a new formulation of controlled diversity, where users explore structured image galleries through meaningful semantic variations.

Bio:

Omer Dahary is a PhD student in Computer Science at Tel Aviv University, advised by Daniel Cohen-Or. His research focuses on generative models, with an emphasis on improving their ability to align with user control.