Deriving insights from experimentally generated datasets requires methods that can account for random and systematic measurement errors and remove them in order to accurately represent the underlying effects of the conditions being tested. Here we present a framework for pretraining on large-scale microscopy datasets that includes three steps: (1) curating a set of diverse and self-consistent training samples, (2) scaling training of an appropriate foundation model architecture on this dataset, (3) evaluating intermediate layers of the trained model to identify the best representation for downstream tasks. Using this strategy, we present the largest foundation model for cell microscopy data to our knowledge, a new 1.9 billion-parameter ViT-G/8 MAE trained on over 8 billion microscopy image crops. Compared to a previous published ViT-L/8 MAE, our new model achieves a 60% improvement in linear separability of genetic perturbations and obtains the best overall performance on whole-genome relationship recall, batch correction replicate consistency, and compound-gene activity prediction benchmarks.

The continued scaling of genetic perturbation technologies combined with high-dimensional assays such as cellular microscopy and RNA-sequencing has enabled genome-scale reverse-genetics experiments that go beyond single-endpoint measurements of growth or lethality. Datasets emerging from these experiments can be combined to construct perturbative "maps of biology", in which readouts from various manipulations (e.g., CRISPR-Cas9 knockout, CRISPRi knockdown, compound treatment) are placed in unified, relatable embedding spaces allowing for the generation of genome-scale sets of pairwise comparisons. These maps of biology capture known biological relationships and uncover new associations which can be used for downstream discovery tasks. Construction of these maps involves many technical choices in both experimental and computational protocols, motivating the design of benchmark procedures to evaluate map quality in a systematic, unbiased manner. Here, we (1) establish a standardized terminology for the steps involved in perturbative map building, (2) introduce key classes of benchmarks to assess the quality of such maps, (3) construct 18 maps from four genome-scale datasets employing different cell types, perturbation technologies, and data readout modalities, (4) generate benchmark metrics for the constructed maps and investigate the reasons for performance variations, and (5) demonstrate utility of these maps to discover new biology by suggesting roles for two largely uncharacterized genes.

Distribution shifts—where the training distribution differs from the test distribution—can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. The full paper, code, and leaderboards are available at https://wilds.stanford.edu.

Biomolecular simulation is a core application on supercomputers, but it is exceptionally difficult to achieve the strong scaling necessary to reach biologically relevant timescales. Here, we present a new paradigm for parallel adaptive molecular dynamics and a publicly available implementation: Copernicus. This framework combines performance-leading molecular dynamics parallelized on three levels (SIMD, threads, and message-passing) with kinetic clustering, statistical model building and real-time result monitoring. Copernicus enables execution as single parallel jobs with automatic resource allocation. Even for a small protein such as villin (9,864 atoms), Copernicus exhibits near-linear strong scaling from 1 to 5,376 AMD cores. Starting from extended chains we observe structures 0.6 Å from the native state within 30h, and achieve sufficient sampling to predict the native state without a priori knowledge after 80--90h. To match Copernicus' efficiency, a classical simulation would have to exceed 50 microseconds per day, currently infeasible even with custom hardware designed for simulations.

Graphics processing units (GPUs) are gaining widespread use in high-performance computing because of their performance advantages relative to CPUs. However, the reliability of GPUs is largely unproven. In particular, current GPUs lack error checking and correcting (ECC) in their memory subsystems. The impact of this design has not been previously measured at a large enough scale to quantify soft error events. We present MemtestG80, our software for assessing memory error rates on NVIDIA graphics cards. Furthermore, we present a large-scale assessment of GPU error rate, conducted by running MemtestG80 on over 50,000 hosts on the Folding@home distributed computing network. Our control experiments on consumer-grade and dedicated-GPGPU hardware in a controlled environment found no errors. However, our survey on Folding@home finds that, in their installed environments, two-thirds of tested GPUs exhibit a detectable, pattern-sensitive rate of memory soft errors. We show that these errors persist after controlling for over clocking and environmental proxies for temperature, but depend strongly on board architecture.