Search for: All records

This study presents an enhanced protein design algorithm that aims to emulate natural heterogeneity of protein sequences. Initial analysis revealed that natural proteins exhibit a permutation composition lower than the theoretical maximum, suggesting a selective utilization of the 20-letter amino acid alphabet. By not constraining the amino acid composition of the protein sequence but instead allowing random reshuffling of the composition, the resulting design algorithm generates sequences that maintain lower permutation compositions in equilibrium, aligning closely with natural proteins. Folding free energy computations demonstrated that the designed sequences refold to their native structures with high precision, except for proteins with large disordered regions. In addition, direct coupling analysis showed a strong correlation between predicted and actual protein contacts, with accuracy exceeding 82% for a large number of top pairs (>4L). The algorithm also resolved biases in previous designs, ensuring a more accurate representation of protein interactions. Overall, it not only mimics the natural heterogeneity of proteins but also ensures correct folding, marking a significant advancement in protein design and engineering. 

Machine learning has been proposed as an alternative to theoretical modeling when dealing with complex problems in biological physics. However, in this perspective, we argue that a more successful approach is a proper combination of these two methodologies. We discuss how ideas coming from physical modeling neuronal processing led to early formulations of computational neural networks, e.g., Hopfield networks. We then show how modern learning approaches like Potts models, Boltzmann machines, and the transformer architecture are related to each other, specifically, through a shared energy representation. We summarize recent efforts to establish these connections and provide examples on how each of these formulations integrating physical modeling and machine learning have been successful in tackling recent problems in biomolecular structure, dynamics, function, evolution, and design. Instances include protein structure prediction; improvement in computational complexity and accuracy of molecular dynamics simulations; better inference of the effects of mutations in proteins leading to improved evolutionary modeling and finally how machine learning is revolutionizing protein engineering and design. Going beyond naturally existing protein sequences, a connection to protein design is discussed where synthetic sequences are able to fold to naturally occurring motifs driven by a model rooted in physical principles. We show that this model is “learnable” and propose its future use in the generation of unique sequences that can fold into a target structure. 

Protein evolution is guided by structural, functional, and dynamical constraints ensuring organismal viability. Pseudogenes are genomic sequences identified in many eukaryotes that lack translational activity due to sequence degradation and thus over time have undergone “devolution.” Previously pseudogenized genes sometimes regain their protein-coding function, suggesting they may still encode robust folding energy landscapes despite multiple mutations. We study both the physical folding landscapes of protein sequences corresponding to human pseudogenes using the Associative Memory, Water Mediated, Structure and Energy Model, and the evolutionary energy landscapes obtained using direct coupling analysis (DCA) on their parent protein families. We found that generally mutations that have occurred in pseudogene sequences have disrupted their native global network of stabilizing residue interactions, making it harder for them to fold if they were translated. In some cases, however, energetic frustration has apparently decreased when the functional constraints were removed. We analyzed this unexpected situation for Cyclophilin A, Profilin-1, and Small Ubiquitin-like Modifier 2 Protein. Our analysis reveals that when such mutations in the pseudogene ultimately stabilize folding, at the same time, they likely alter the pseudogenes’ former biological activity, as estimated by DCA. We localize most of these stabilizing mutations generally to normally frustrated regions required for binding to other partners. 

Abstract Variational autoencoders are unsupervised learning models with generative capabilities, when applied to protein data, they classify sequences by phylogeny and generate de novo sequences which preserve statistical properties of protein composition. While previous studies focus on clustering and generative features, here, we evaluate the underlying latent manifold in which sequence information is embedded. To investigate properties of the latent manifold, we utilize direct coupling analysis and a Potts Hamiltonian model to construct a latent generative landscape. We showcase how this landscape captures phylogenetic groupings, functional and fitness properties of several systems including Globins,β-lactamases, ion channels, and transcription factors. We provide support on how the landscape helps us understand the effects of sequence variability observed in experimental data and provides insights on directed and natural protein evolution. We propose that combining generative properties and functional predictive power of variational autoencoders and coevolutionary analysis could be beneficial in applications for protein engineering and design. 

Computational models of evolution are valuable for understanding the dynamics of sequence variation, to infer phylogenetic relationships or potential evolutionary pathways and for biomedical and industrial applications. Despite these benefits, few have validated their propensities to generate outputs with in vivo functionality, which would enhance their value as accurate and interpretable evolutionary algorithms. We demonstrate the power of epistasis inferred from natural protein families to evolve sequence variants in an algorithm we developed called sequence evolution with epistatic contributions (SEEC). Utilizing the Hamiltonian of the joint probability of sequences in the family as fitness metric, we sampled and experimentally tested for in vivo $\beta$-lactamase activity inEscherichia coliTEM-1 variants. These evolved proteins can have dozens of mutations dispersed across the structure while preserving sites essential for both catalysis and interactions. Remarkably, these variants retain family-like functionality while being more active than their wild-type predecessor. We found that depending on the inference method used to generate the epistatic constraints, different parameters simulate diverse selection strengths. Under weaker selection, local Hamiltonian fluctuations reliably predict relative changes to variant fitness, recapitulating neutral evolution. SEEC has the potential to explore the dynamics of neofunctionalization, characterize viral fitness landscapes, and facilitate vaccine development. 

Mitochondria play a central role in muscle metabolism and function. A unique family of iron–sulfur proteins, termed CDGSH Iron Sulfur Domain-containing (CISD/NEET) proteins, support mitochondrial function in skeletal muscles. The abundance of these proteins declines during aging leading to muscle degeneration. Although the function of the outer mitochondrial CISD/NEET proteins, CISD1/mitoNEET and CISD2/NAF-1, has been defined in skeletal muscle cells, the role of the inner mitochondrial CISD protein, CISD3/MiNT, is currently unknown. Here, we show that CISD3 deficiency in mice results in muscle atrophy that shares proteomic features with Duchenne muscular dystrophy. We further reveal that CISD3 deficiency impairs the function and structure of skeletal muscles, as well as their mitochondria, and that CISD3 interacts with, and donates its [2Fe-2S] clusters to, complex I respiratory chain subunit NADH Ubiquinone Oxidoreductase Core Subunit V2 (NDUFV2). Using coevolutionary and structural computational tools, we model a CISD3–NDUFV2 complex with proximal coevolving residue interactions conducive of [2Fe-2S] cluster transfer reactions, placing the clusters of the two proteins 10 to 16 Å apart. Taken together, our findings reveal that CISD3/MiNT is important for supporting the biogenesis and function of complex I, essential for muscle maintenance and function. Interventions that target CISD3 could therefore impact different muscle degeneration syndromes, aging, and related conditions.