Search for: All records

AI-based frameworks for protein engineering use self-supervised learning (SSL) to obtain representations for downstream biological predictions. The most common training objective for these methods is wildtype accuracy: given a sequence or structure where a wildtype residue has been masked, predict the missing amino acid. Wildtype accuracy, however, does not align with the primary goal of protein engineering, which is to suggest a {\em mutation} rather than to identify what already appears in nature. Here we present Evolutionary Ranking (EvoRank), a training objective that incorporates evolutionary information derived from multiple sequence alignments (MSAs) to learn more diverse protein representations. EvoRank corresponds to ranking amino-acid likelihoods in the probability distribution induced by an MSA. This objective forces models to learn the underlying evolutionary dynamics of a protein. Across a variety of phenotypes and datasets, we demonstrate that EvoRank leads to dramatic improvements in zero-shot performance and can compete with models fine-tuned on experimental data. This is particularly important in protein engineering, where it is expensive to obtain data for fine-tuning. 

We employ two machine learning techniques, i.e., neural networks and genetic-programming-based symbolic regression, to examine the dynamics of the electron-positron pair creation process with full space–time resolution inside the interaction zone of a supercritical electric field pulse. Both algorithms receive multiple sequences of partially dressed electronic and positronic spatial probability densities as training data and exploit their features as a function of the dressing strength in order to predict each particle’s spatial distribution inside the electric field. A linear combination of both predicted densities is then compared with the unambiguous total charge density, which also contains contributions associated with the independent vacuum polarization process. After its subtraction, the good match confirms the validity of the machine learning approach and lends some credibility to the validity of the predicted single-particle densities. 

We examine the effect of a frequency-chirped external force field on the final energy that has been absorbed by two classical mechanical oscillators, by quantum mechanical two- and three-level systems, and by electron-positron pairs that were created from the quantum field theoretical Dirac vacuum. By comparing the final dynamical responses to the original force field with that associated with the corresponding time-reversed field, we can test the sensitivity of each of these five systems to the temporal phase information contained in the field. We predict that the linear oscillator, the two-level atom, and the pair-creation process triggered by a spatially homogeneous field are remarkably immune to this phase, whereas the quartic oscillator, the three-level atom, or the pair-creation process caused by a space-time field absorb the provided energy differently depending on the temporal details of the external field.