Machine learning interatomic potentials (MLIPs) have become increasingly effective at approximating quantum mechanical calculations at a fraction of the computational cost. However, lower errors on held out test sets do not always translate to improved results on downstream physical property prediction tasks. In this paper, we propose testing MLIPs on their practical ability to conserve energy during molecular dynamic simulations. If passed, improved correlations are found between test errors and their performance on physical property prediction tasks. We identify choices which may lead to models failing this test, and use these observations to improve upon highly-expressive models. The resulting model, eSEN, provides state-of-the-art results on a range of physical property prediction tasks, including materials stability prediction, thermal conductivity prediction, and phonon calculations.

The ability to quickly and accurately compute properties from atomic simulations is critical for advancing a large number of applications in chemistry and materials science including drug discovery, energy storage, and semiconductor manufacturing. To address this need, we present a family of Universal Models for Atoms (UMA), designed to push the frontier of speed, accuracy, and generalization. UMA models are trained on half a billion unique 3D atomic structures (the largest training runs to date) by compiling data across multiple chemical domains, e.g. molecules, materials, and catalysts. We develop empirical scaling laws to help understand how to increase model capacity alongside dataset size to achieve the best accuracy. The UMA small and medium models utilize a novel architectural design we refer to as mixture of linear experts that enables increasing model capacity without sacrificing speed. For example, UMA-medium has 1.4B parameters but only $\sim$50M active parameters per atomic structure. We evaluate UMA models on a diverse set of applications across multiple domains and find that, remarkably, a single model without any fine-tuning can perform similarly or better than specialized models. We are releasing the UMA code, weights, and associated data to accelerate computational workflows and enable the community to build increasingly capable AI models.

Even though neural networks enjoy widespread use, they still struggle to learn the basic laws of physics. How might we endow them with better inductive biases? In this paper, we draw inspiration from Hamiltonian mechanics to train models that learn and respect exact conservation laws in an unsupervised manner. We evaluate our models on problems where conservation of energy is important, including the two-body problem and pixel observations of a pendulum. Our model trains faster and generalizes better than a regular neural network. An interesting side effect is that our model is perfectly reversible in time.

Abstract Summary: We report on a major update (version 2) of the original SHort Read Mapping Program (SHRiMP). SHRiMP2 primarily targets mapping sensitivity, and is able to achieve high accuracy at a very reasonable speed. SHRiMP2 supports both letter space and color space (AB/SOLiD) reads, enables for direct alignment of paired reads and uses parallel computation to fully utilize multi-core architectures. Availability: SHRiMP2 executables and source code are freely available at: http://compbio.cs.toronto.edu/shrimp/. Contact: [email protected] Supplementary information: Supplementary data are available at Bioinformatics online.

Background Early detection of autism can improve the quality of life of affected individuals [1]. Qualitative screening methods continue to improve, but still suffer from low sensitivity despite increasing specificity [2,3]. In collaboration with the Hospital for Sick Children, we are sequencing the exomes of 1 000 individuals with autism in order to discover genetic variants associated with the disorder. Discovery of associated variants can lead to earlier diagnosis and treatment.