Enhancing efficiency of protein language models with minimal wet-lab data through few-shot learning.

Accurately modeling the protein fitness landscapes holds great importance for protein engineering. Pre-trained protein language models have achieved state-of-the-art performance in predicting protein fitness without wet-lab experimental data, but their accuracy and interpretability remain limited. On the other hand, traditional supervised deep learning models require abundant labeled training examples for performance improvements, posing a practical barrier. In this work, we introduce FSFP, a training strategy that can effectively optimize protein language models under extreme data scarcity for fitness prediction. By combining meta-transfer learning, learning to rank, and parameter-efficient fine-tuning, FSFP can significantly boost the performance of various protein language models using merely tens of labeled single-site mutants from the target protein. In silico benchmarks across 87 deep mutational scanning datasets demonstrate FSFP's superiority over both unsupervised and supervised baselines. Furthermore, we successfully apply FSFP to engineer the Phi29 DNA polymerase through wet-lab experiments, achieving a 25% increase in the positive rate. These results underscore the potential of our approach in aiding AI-guided protein engineering.

The evolution of anionic nanoclusters at the electrode interface in water-in-salt electrolytes.

Water-in-salt electrolytes (WiSEs) have attracted extensive attention as promising alternatives to organic electrolytes. The limited electrochemical stability windows (ESWs) of aqueous electrolytes are significantly widened by WiSEs. However, the actual ESWs are lower than predicted as the interphase with WiSEs is not as stable as the solid electrolyte interphase (SEI) in conventional lithium-ion batteries. Therefore, identifying the interface state in WiSEs is vital to understanding their electrochemical behavior. Here, the structure of the lithium bis(trifluoromethane sulfonyl)imide (LiTFSI) electrolyte near the interface of the carbon electrode (Ketjen black) was evaluated by experimental methods (neutron diffraction, Raman, and nuclear magnetic resonance spectroscopy) and molecular dynamics (MD) simulations. The results reveal that the introduction of carbon electrodes increases the size of the anionic nanoclusters and enhances the microphase separation at the interface. The MD simulations show that cation-π interactions are responsible for the evolution of anionic nanoclusters at the electrode interface. Moreover, lower charge transfer resistance is achieved at carbon-based electrodes due to the specific interface state. Our findings provide a strategy for introducing cation-π interactions between electrodes and electrolytes to improve the electrochemical performance.

The ω3 scaling of the vibrational density of states in quasi-2D nanoconfined solids.

The vibrational properties of crystalline bulk materials are well described by Debye theory, which successfully predicts the quadratic ω2 low-frequency scaling of the vibrational density of states. However, the analogous framework for nanoconfined materials with fewer degrees of freedom has been far less well explored. Using inelastic neutron scattering, we characterize the vibrational density of states of amorphous ice confined inside graphene oxide membranes and we observe a crossover from the Debye ω2 scaling to an anomalous ω3 behaviour upon reducing the confinement size L. Additionally, using molecular dynamics simulations, we confirm the experimental findings and prove that such a scaling appears in both crystalline and amorphous solids under slab-confinement. We theoretically demonstrate that this low-frequency ω3 law results from the geometric constraints on the momentum phase space induced by confinement along one spatial direction. Finally, we predict that the Debye scaling reappears at a characteristic frequency ω× = vL/2π, with v the speed of sound of the material, and we confirm this quantitative estimate with simulations.