8 October 2021 in Science
A biological clock in a test tube: The biological clock of cyanobacteria, which remarkably requires just three proteins, has been reconstituted in vitro in a system that allows detailed study of its inputs and outputs, bringing new understanding of how environmental signals can influence a biological oscillator and how the clock controls cellular events such as gene transcription. Chavan et al . extended the known in vitro function of the core clock components to include output signals to transcriptional regulation and allow monitoring through fluorescence measurements in real time. The authors combined crystallography, mutagenesis, and quantitative modeling to further explore the clock mechanism, which may enable future synthetic biology applications. —LBR
6 October 2026 in bioRxiv
The
A
TPases
A
ssociated with various cellular
A
ctivities (AAA) superfamily utilizes a conserved three-dimensional structure to drive diverse mechanical workflows. While cryo-electron microscopy has revealed extensive structural data, predicting how specific sequence variations dictate unique biochemical mechanisms remains a challenge. Here, we present a machine learning framework to resolve these sequence-function relationships by mapping the functional landscape of AAA proteins. We established a hierarchical classification scheme based on five distinct modes of coupling ATP hydrolysis to mechanical work. Using CatFunc , a lightweight protein language model (PLM) pipeline, we leveraged evolutionary context from proteomic sequences for supervised classification of these mechanisms. Excellent recall of our mechanistic annotations across the AAA hierarchy, demonstrating that mechanistic definitions are encapsulated at the domain level. Notably, systematic misclassifications revealed a shared evolutionary and mechanistic ancestry within the rotary motor class. Applying the model to out-of-distribution proteomic datasets, we established biophysical metrics to detect unrepresented functions, demonstrating that restricting inference to isolated AAA domains significantly improves classification accuracy for novel sequences. Our framework provides a scalable strategy for decoding functional phenotypes from sequence data, offering a transferable approach for other biomedically vital families like G protein-coupled receptors and kinases.