The power and pitfalls of AlphaFold2 for structure prediction beyond rigid globular proteins.

Artificial intelligence-driven advances in protein structure prediction in recent years have raised the question: has the protein structure-prediction problem been solved? Here, with a focus on nonglobular proteins, we highlight the many strengths and potential weaknesses of DeepMind's AlphaFold2 in the context of its biological and therapeutic applications. We summarize the subtleties associated with evaluation of AlphaFold2 model quality and reliability using the predicted local distance difference test (pLDDT) and predicted aligned error (PAE) values. We highlight various classes of proteins that AlphaFold2 can be applied to and the caveats involved. Concrete examples of how AlphaFold2 models can be integrated with experimental data in the form of small-angle X-ray scattering (SAXS), solution NMR, cryo-electron microscopy (cryo-EM) and X-ray diffraction are discussed. Finally, we highlight the need to move beyond structure prediction of rigid, static structural snapshots toward conformational ensembles and alternate biologically relevant states. The overarching theme is that careful consideration is due when using AlphaFold2-generated models to generate testable hypotheses and structural models, rather than treating predicted models as de facto ground truth structures.

Structural Consequences of Introducing Bioactive Domains to Designer ß-Sheet Peptide Self-Assemblies.

We applied solid- and solution-state nuclear magnetic resonance spectroscopy to examine the structure of multidomain peptides composed of self-assembling ß-sheet domains linked to bioactive domains. Bioactive domains can be selected to stimulate specific biological responses (e.g., via receptor binding), while the ß-sheets provide the desirable nanoscale properties. Although previous work has established the efficacy of multidomain peptides, molecular-level characterization is lacking. The bioactive domains are intended to remain solvent-accessible without being incorporated into the ß-sheet structure. We tested for three possible anticipated molecular-level consequences of introducing bioactive domains to ß-sheet-forming peptides: (1) the bioactive domain has no effect on the self-assembling peptide structure; (2) the bioactive domain is incorporated into the ß-sheet nanofiber; and (3) the bioactive domain interferes with self-assembly such that nanofibers are not formed. The peptides involved in this study incorporated self-assembling domains based on the (SL)6 motif and bioactive domains including a VEGF-A mimic (QK), an IGF-mimic (IGF-1c), and a de novo SARS-CoV-2 binding peptide (SBP3). We observed all three of the anticipated outcomes from our examination of peptides, illustrating the unintended structural effects that could adversely affect the desired biofunctionality and biomaterial properties of the resulting peptide hydrogel. This work is the first attempt to evaluate the structural effects of incorporating bioactive domains into a set of peptides unified by a similar self-assembling peptide domain. These structural insights reveal unmet challenges in the design of highly tunable bioactive self-assembling peptide hydrogels.

Disordered regions in proteusin peptides guide post-translational modification by a flavin-dependent RiPP brominase.

To biosynthesize ribosomally synthesized and post-translationally modified peptides (RiPPs), enzymes recognize and bind to the N-terminal leader region of substrate peptides which enables catalytic modification of the C-terminal core. Our current understanding of RiPP leaders is that they are short and largely unstructured. Proteusins are RiPP precursor peptides that defy this characterization as they possess unusually long leaders. Proteusin peptides have not been structurally characterized, and we possess scant understanding of how these atypical leaders engage with modifying enzymes. Here, we determine the structure of a proteusin peptide which shows that unlike other RiPP leaders, proteusin leaders are preorganized into a rigidly structured region and a smaller intrinsically disordered region. With residue level resolution gained from NMR titration experiments, the intermolecular peptide-protein interactions between proteusin leaders and a flavin-dependent brominase are mapped onto the disordered region, leaving the rigidly structured region of the proteusin leader to be functionally dispensable. Spectroscopic observations are biochemically validated to identify a binding motif in proteusin peptides that is conserved among other RiPP leaders as well. This study provides a structural characterization of the proteusin peptides and extends the paradigm of RiPP modification enzymes using not only unstructured peptides, but also structured proteins as substrates.

Generative ß-hairpin design using a residue-based physicochemical property landscape.

De novo peptide design is a new frontier that has broad application potential in the biological and biomedical fields. Most existing models for de novo peptide design are largely based on sequence homology that can be restricted based on evolutionarily derived protein sequences and lack the physicochemical context essential in protein folding. Generative machine learning for de novo peptide design is a promising way to synthesize theoretical data that are based on, but unique from, the observable universe. In this study, we created and tested a custom peptide generative adversarial network intended to design peptide sequences that can fold into the ß-hairpin secondary structure. This deep neural network model is designed to establish a preliminary foundation of the generative approach based on physicochemical and conformational properties of 20 canonical amino acids, for example, hydrophobicity and residue volume, using extant structure-specific sequence data from the PDB. The beta generative adversarial network model robustly distinguishes secondary structures of ß hairpin from α helix and intrinsically disordered peptides with an accuracy of up to 96% and generates artificial ß-hairpin peptide sequences with minimum sequence identities around 31% and 50% when compared against the current NCBI PDB and nonredundant databases, respectively. These results highlight the potential of generative models specifically anchored by physicochemical and conformational property features of amino acids to expand the sequence-to-structure landscape of proteins beyond evolutionary limits.

Antiviral fibrils of self-assembled peptides with tunable compositions.

The lasting threat of viral pandemics necessitates the development of tailorable first-response antivirals with specific but adaptive architectures for treatment of novel viral infections. Here, such an antiviral platform has been developed based on a mixture of hetero-peptides self-assembled into functionalized ß-sheets capable of specific multivalent binding to viral protein complexes. One domain of each hetero-peptide is designed to specifically bind to certain viral proteins, while another domain self-assembles into fibrils with epitope binding characteristics determined by the types of peptides and their molar fractions. The self-assembled fibrils maintain enhanced binding to viral protein complexes and retain high resilience to viral mutations. This method is experimentally and computationally tested using short peptides that specifically bind to Spike proteins of SARS-CoV-2. This platform is efficacious, inexpensive, and stable with excellent tolerability.

In silico identification and characterization of small molecule binding to the CD1d immunoreceptor.

CD1 immunoreceptors are a non-classical major histocompatibility complex (MHC) that present antigens to T cells to elucidate immune responses against disease. The antigen repertoire of CD1 has been composed primarily of lipids until recently when CD1d-restricted T cells were shown to be activated by non-lipidic small molecules, such as phenyl pentamethyl dihydrobenzofuran sulfonate (PPBF) and related benzofuran sulfonates. To date structural insights into PPBF/CD1d interactions are lacking, so it is unknown whether small molecule and lipid antigens are presented and recognized through similar mechanisms. Furthermore, it is unknown whether CD1d can bind to and present a broader range of small molecule metabolites to T cells, acting out functions analogous to the MHC class I related protein MR1. Here, we perform in silico docking and molecular dynamics simulations to structurally characterize small molecule interactions with CD1d. PPBF was supported to be presented to T cell receptors through the CD1d F' pocket. Virtual screening of CD1d against more than 17,000 small molecules with diverse geometry and chemistry identified several novel scaffolds, including phytosterols, cholesterols, triterpenes, and carbazole alkaloids, that serve as candidate CD1d antigens. Protein-ligand interaction profiling revealed conserved residues in the CD1d F' pocket that similarly anchor small molecules and lipids. Our results suggest that CD1d could have the intrinsic ability to bind and present a broad range of small molecule metabolites to T cells to carry out its function beyond lipid antigen presentation.Communicated by Ramaswamy H. Sarma.