The ALCF user community continually pushes the boundaries of scientific computing, producing groundbreaking studies in areas ranging from chemistry and engineering to physics and materials science. In this regular feature, we highlight some of the recent results published by ALCF users.

“Computational Methods for Peptide Macrocycle Drug Design,” Peptide Therapeutics: Fundamentals of Design, Development, and Delivery

ALCF principal investigator (PI): Vikram Mulligan, Flatiron Institute

Peptide macrocycles represent a promising class of therapeutics, albeit one that is under-represented amongst existing drugs. One disadvantage of macrocycles, however, is that they can be more conformationally heterogeneous than small molecules or large, well-folded proteins. This flexibility can impede high-affinity binding. In recent years, the development of new computational tools has made possible the structure-based design of macrocycles that are able to fold into rigid structures compatible with binding to target proteins of therapeutic interest. This chapter, written as part of an INCITE allocation, is intended to introduce biologists, chemists, and drug developers to current computational methods for peptide macrocycle drug design.

“Mining for Strong Gravitational Lenses with Self-Supervised Learning,” The Astrophysical Journal

ALCF PI: George Stein, University of California Berkeley

The authors employ ALCF resources in their work to employ self-supervised representation learning to distill information from 76 million galaxy images from the Dark Energy Spectroscopic Instrument Legacy Imaging Surveys’ Data Release 9. Targeting the identification of new strong gravitational lens candidates, the authors first create a rapid similarity search tool to discover new strong lenses given only a single labeled example. They show how training a simple linear classifier on the self-supervised representations can automatically classify strong lenses with great efficiency, and then present 1192 new strong lens candidates identified through a brief campaign.

“A Kinetic Energy- and Entropy-Preserving Scheme for Compressible Two-Phase Flows,” Journal of Computational Physics

ALCF PI: Parviz Moin

Accurate numerical modeling of compressible flows, particularly in the turbulent regime, requires a method that is non-dissipative and stable at high Reynolds numbers. For a compressible flow, it is known that discrete conservation of kinetic energy is not a sufficient condition for numerical stability, unlike in incompressible flows. The authors, utilizing ALCF systems via an ALCC allocation, adopt the recently developed conservative diffuse-interface method for the simulation of compressible two-phase flows.

“Adaptive Determination of the Optimal Exchange Location in Wall-Modeled Large-Eddy Simulation,” AIAA Journal

ALCF PI: Johann Larsson, University of Maryland

Wall-modeled large-eddy simulations introduce a modeling interface (or exchange location) separating the wall-modeled layer from the rest of the domain. The current state-of-the-art is to rely on user expertise when choosing where to place this modeling interface, whether this choice is tied to the grid or not. This paper presents a postprocessing algorithm—developed with ALCF computing resources—that determines the exchange location systematically.

 “Modular Method for Estimation of Velocity and Temperature Profiles in High-Speed Boundary Layers,” AIAA Journal

ALCF PI: Johann Larsson, University of Maryland

The fluid property variation caused by viscous heating affects the mean velocity and Reynolds stresses in compressible turbulent boundary layers, and as a result also affects the resulting skin friction and wall heat transfer. Developed using ALCF computing time, the authors present a method that estimates the mean velocity and temperature profiles, and therefore also the friction and heat transfer coefficients, for a given Mach number, Reynolds number, and wall thermal condition.