Large language models generate functional protein sequences across diverse families.

Deep-learning language models have shown promise in various biotechnological applications, including protein design and engineering. Here we describe ProGen, a language model that can generate protein sequences with a predictable function across large protein families, akin to generating grammatically and semantically correct natural language sentences on diverse topics. The model was trained on 280 million protein sequences from >19,000 families and is augmented with control tags specifying protein properties. ProGen can be further fine-tuned to curated sequences and tags to improve controllable generation performance of proteins from families with sufficient homologous samples. Artificial proteins fine-tuned to five distinct lysozyme families showed similar catalytic efficiencies as natural lysozymes, with sequence identity to natural proteins as low as 31.4%. ProGen is readily adapted to diverse protein families, as we demonstrate with chorismate mutase and malate dehydrogenase.

Elucidating the potential of crude cell extracts for producing pyruvate from glucose.

Living systems possess a rich biochemistry that can be harnessed through metabolic engineering to produce valuable therapeutics, fuels and fine chemicals. In spite of the tools created for this purpose, many organisms tend to be recalcitrant to modification or difficult to optimize. Crude cellular extracts, made by lysis of cells, possess much of the same biochemical capability, but in an easier to manipulate context. Metabolic engineering in crude extracts, or cell-free metabolic engineering, can harness these capabilities to feed heterologous pathways for metabolite production and serve as a platform for pathway optimization. However, the inherent biochemical potential of a crude extract remains ill-defined, and consequently, the use of such extracts can result in inefficient processes and unintended side products. Herein, we show that changes in cell growth conditions lead to changes in the enzymatic activity of crude cell extracts and result in different abilities to produce the central biochemical precursor pyruvate when fed glucose. Proteomic analyses coupled with metabolite measurements uncover the diverse biochemical capabilities of these different crude extract preparations and provide a framework for how analytical measurements can be used to inform and improve crude extract performance. Such informed developments can allow enrichment of crude extracts with pathways that promote or deplete particular metabolic processes and aid in the metabolic engineering of defined products.