Biomimetic Synthesis of Nanosilica by Deep Learning‐Designed Peptides and Its Anti‐UV Application

Abstract

The silicifying peptide R5 (H‐SSKKSGSYSGSKGSKRRIL‐OH) derived from diatoms is extensively investigated, but the mechanism underlying silica synthesis by R5 or R5‐alike peptides is still poorly understood, limiting the design of silicifying peptides Herein, machine learning techniques are used to design peptides with silicifying functionality. Utilizing a comprehensive dataset of peptides and their corresponding silicification outcomes, a deep learning model based on antimicrobial peptide migration learning is created. This model exhibits the remarkable capability to accurately predict peptide sequences with a high potential for facilitating silica formation. A selection of artificially designed peptides can catalyze the biomimetic synthesis of nanosilica, some of which demonstrate better catalytic activity with a wider pH range and faster reaction rate compared with the peptide R5. Additionally, the designed peptide is used to wrap the model diatom Phaeodactylum tricornutum with nanosilica coatings, resulting in a significant enhancement in the UV resistance of cells. The new silicified peptides are highly significant for advancing the understanding of the silica synthesis mechanism in diatoms, and the encapsulation of P. tricornutum has potential benefits in the development of new biosensors.