Search

Translation is an essential attribute of all living cells. At the heart of cellular operation, it is a chemical information decoding process that begins with an input string of nucleotides and ends with the synthesis of a specific output string of peptides. The translation process is interconnected with gene expression, physiological regulation, transcription, and responses to signaling molecules, among other cellular functions. Foundational efforts have uncovered a wealth of knowledge about the mechanistic functions of the components of translation and their many interactions between them, but the broader biochemical connections between translation, metabolism and polymer biosynthesis that enable translation to occur have not been comprehensively mapped. Here we present a multilayer graph of biochemical reactions describing the translation, polymer biosynthesis and metabolism networks of an Escherichia coli cell. Intriguingly, the compounds that compose these three layers are distinctly aggregated into three modes regardless of their layer categorization. Multimodal mass distributions are well-known in ecosystems, but this is the first such distribution reported at the biochemical level. The degree distributions of the translation and metabolic networks are each likely to be heavy-tailed, but the polymer biosynthesis network is not. A multimodal mass-degree distribution indicates that the translation and metabolism networks are each distinct, adaptive biochemical modules, and that the gaps between the modes reflect evolved responses to the functional use of metabolite, polypeptide and polynucleotide compounds. The chemical reaction network of cellular translation opens new avenues for exploring complex adaptive phenomena such as percolation and phase changes in biochemical contexts.

Machine learning milestones in computational chemistry are overshadowed by their unaccountability and the overwhelming zoo of tools for each specific task. A promising path to tackle these problems is using machine learning to reproduce physical magnitudes as a basis to derive many other properties. By using a model of the electron density consisting of an analytical expansion on a linear set of isotropic and anisotropic functions, we implemented in this work a message-passing neural network able to reproduce electron density in molecules with just a 2.5% absolute error in complex cases. We also adapted our methodology to describe electron density in large biomolecules (proteins) and to obtain atomic charges, interaction energies, and DFT energies. We show that electron density learning is a new promising avenue with a variety of forthcoming applications.

CD1 molecules present lipid antigens for recognition by T-cell receptors (TCRs). Although a reasonably detailed picture of the CD1-lipid-TCR interaction exists, the initial steps regarding lipid loading onto and exchange between CD1 proteins remain elusive. The hydrophobic nature of lipids and the fact that CD1 molecules are unable to extract lipids from membranes raise the need for the assistance of helper proteins in lipid trafficking. However, the experimental study of this traffic in the endosomal compartments at which it occurs is so challenging that computational studies can help provide mechanistic insight into the associated processes. Here we present a multifaceted computational approach to obtain dynamic structural data on the human CD1d isotype. Conformational dynamics analysis shows an intrinsic flexibility associated with the protein architecture. Electrostatic properties together with molecular dynamics results for CD1d complexes with several lipids and helper proteins unravel the high dynamic plasticity of the antigen-binding site that is crucially favoured by acidic pH and the presence of helper proteins.

The remarkable ability of tardigrades to withstand a wide range of physical and chemical extremes has attracted a considerable interest in these small invertebrates, with a particular focus on the protective roles of proteins expressed during such conditions. The discovery that a tardigrade-unique protein named Dsup (damage suppressor) protects DNA from damage produced by radiation and radicals, has raised expectations concerning its potential applications in biotechnology and medicine. We present in this paper what might be dubbed a "computational experiment" on the Dsup-DNA system. By means of molecular modelling, calculations of electrostatic potentials and electric fields, and all-atom molecular dynamics simulations, we obtained a dynamic picture of the Dsup-DNA interaction. Our results suggest that the protein is intrinsically disordered, which enables Dsup to adjust its structure to fit DNA shape. Strong electrostatic attractions and high protein flexibility drive the formation of a molecular aggregate in which Dsup shields DNA. While the precise mechanism of DNA protection conferred by Dsup remains to be elucidated, our study provides some molecular clues of their association that could be of interest for further investigation in this line.

In nitrogenase biosynthesis, the iron-molybdenum cofactor (FeMo-co) is externally assembled at scaffold proteins and delivered to the NifDK nitrogenase component by the NafY metallochaperone. Here we have used nuclear magnetic resonance, molecular dynamics, and functional analysis to elucidate the environment and coordination of FeMo-co in NafY. H-121 stands as the key FeMo-co ligand. Regions near FeMo-co diverge from H-121 and include the eta 1, alpha 1, alpha 2 helical lobe and a narrow path between H-121 and C-196.