Exploring the potential of structure-based deep learning approaches for T cell receptor design.

Deep learning methods, trained on the increasing set of available protein 3D structures and sequences, have substantially impacted the protein modeling and design field. These advancements have facilitated the creation of novel proteins, or the optimization of existing ones designed for specific functions, such as binding a target protein. Despite the demonstrated potential of such approaches in designing general protein binders, their application in designing immunotherapeutics remains relatively underexplored. A relevant application is the design of T cell receptors (TCRs). Given the crucial role of T cells in mediating immune responses, redirecting these cells to tumor or infected target cells through the engineering of TCRs has shown promising results in treating diseases, especially cancer. However, the computational design of TCR interactions presents challenges for current physics-based methods, particularly due to the unique natural characteristics of these interfaces, such as low affinity and cross-reactivity. For this reason, in this study, we explored the potential of two structure-based deep learning protein design methods, ProteinMPNN and ESM-IF1, in designing fixed-backbone TCRs for binding target antigenic peptides presented by the MHC through different design scenarios. To evaluate TCR designs, we employed a comprehensive set of sequence- and structure-based metrics, highlighting the benefits of these methods in comparison to classical physics-based design methods and identifying deficiencies for improvement.

Mapping Conformational Changes in the Saliva Proteome Potentially Associated with Oral Cancer Aggressiveness.

Diverse proteomics-based strategies have been applied to saliva to quantitatively identify diagnostic and prognostic targets for oral cancer. Considering that these targets may be regulated by events that do not imply variation in protein abundance levels, we hypothesized that changes in protein conformation can be associated with diagnosis and prognosis, revealing biological processes and novel targets of clinical relevance. For this, we employed limited proteolysis-mass spectrometry in saliva samples to explore structural alterations, comparing the proteome of healthy control and oral squamous cell carcinoma (OSCC) patients with and without lymph node metastasis. Thirty-six proteins with potential structural rearrangements were associated with clinical patient features including transketolase and its interacting partners. Moreover, N-glycosylated peptides contribute to structural rearrangements of potential diagnostic and prognostic markers. Altogether, this approach utilizes saliva proteins to search for targets for diagnosing and prognosing oral cancer and can guide the discovery of potential regulated sites beyond protein-level abundance.

Exploring the Potential of Structure-Based Deep Learning Approaches for T cell Receptor Design.

Deep learning methods, trained on the increasing set of available protein 3D structures and sequences, have substantially impacted the protein modeling and design field. These advancements have facilitated the creation of novel proteins, or the optimization of existing ones designed for specific functions, such as binding a target protein. Despite the demonstrated potential of such approaches in designing general protein binders, their application in designing immunotherapeutics remains relatively unexplored. A relevant application is the design of T cell receptors (TCRs). Given the crucial role of T cells in mediating immune responses, redirecting these cells to tumor or infected target cells through the engineering of TCRs has shown promising results in treating diseases, especially cancer. However, the computational design of TCR interactions presents challenges for current physics-based methods, particularly due to the unique natural characteristics of these interfaces, such as low affinity and cross-reactivity. For this reason, in this study, we explored the potential of two structure-based deep learning protein design methods, ProteinMPNN and ESM-IF, in designing fixed-backbone TCRs for binding target antigenic peptides presented by the MHC through different design scenarios. To evaluate TCR designs, we employed a comprehensive set of sequence- and structure-based metrics, highlighting the benefits of these methods in comparison to classical physics-based design methods and identifying deficiencies for improvement.

Cavity Characterization in Supramolecular Cages.

Confining molecular guests within artificial hosts has provided a major driving force in the rational design of supramolecular cages with tailored properties. Over the last 30 years, a set of design strategies have been developed that enabled the controlled synthesis of a myriad of cages. Recently, there has been a growing interest in involving in silico methods in this toolbox. Cavity shape and size are important parameters that can be easily accessed by inexpensive geometric algorithms. Although these algorithms are well developed for the detection of nonartificial cavities (e.g., enzymes), they are not routinely used for the rational design of supramolecular cages. In order to test the capabilities of this tool, we have evaluated the performance and characteristics of seven different cavity characterization software in the context of 22 analogues of well-known supramolecular cages. Among the tested software, KVFinder project and Fpocket proved to be the most software to characterize supramolecular cavities. With the results of this work, we aim to popularize this underused technique within the supramolecular community.

KVFinder-web: a web-based application for detecting and characterizing biomolecular cavities.

Molecular interactions that modulate catalytic processes occur mainly in cavities throughout the molecular surface. Such interactions occur with specific small molecules due to geometric and physicochemical complementarity with the receptor. In this scenario, we present KVFinder-web, an open-source web-based application of parKVFinder software for cavity detection and characterization of biomolecular structures. The KVFinder-web has two independent components: a RESTful web service and a web graphical portal. Our web service, KVFinder-web service, handles client requests, manages accepted jobs, and performs cavity detection and characterization on accepted jobs. Our graphical web portal, KVFinder-web portal, provides a simple and straightforward page for cavity analysis, which customizes detection parameters, submits jobs to the web service component, and displays cavities and characterizations. We provide a publicly available KVFinder-web at https://kvfinder-web.cnpem.br, running in a cloud environment as docker containers. Further, this deployment type allows KVFinder-web components to be configured locally and customized according to user demand. Hence, users may run jobs on a locally configured service or our public KVFinder-web.

Unsaturated fatty acids as high-affinity ligands of the C-terminal Per-ARNT-Sim domain from the Hypoxia-inducible factor 3α.

Hypoxia-inducible transcription factors (HIF) form heterodimeric complexes that mediate cell responses to hypoxia. The oxygen-dependent stability and activity of the HIF-α subunits is traditionally associated to post-translational modifications such as hydroxylation, acetylation, ubiquitination, and phosphorylation. Here we report novel evidence showing that unsaturated fatty acids are naturally occurring, non-covalent structural ligands of HIF-3α, thus providing the initial framework for exploring its exceptional role as a lipid sensor under hypoxia.