PSSNet-An Accurate Super-Secondary Structure for Protein Segmentation.

A super-secondary structure (SSS) is a spatially unique ensemble of secondary structural elements that determine the three-dimensional shape of a protein and its function, rendering SSSs attractive as folding cores. Understanding known types of SSSs is important for developing a deeper understanding of the mechanisms of protein folding. Here, we propose a universal PSSNet machine-learning method for SSS recognition and segmentation. For various types of SSS segmentation, this method uses key characteristics of SSS geometry, including the lengths of secondary structural elements and the distances between them, torsion angles, spatial positions of Cα atoms, and primary sequences. Using four types of SSSs (ßαß-unit, α-hairpin, ß-hairpin, αα-corner), we showed that extensive SSS sets could be reliably selected from the Protein Data Bank and AlphaFold 2.0 database of protein structures.

3ß-Corner Stability by Comparative Molecular Dynamics Simulations.

This study explored the mechanisms by which the stability of super-secondary structures of the 3ß-corner type autonomously outside the protein globule are maintained in an aqueous environment. A molecular dynamic (MD) study determined the behavioral diversity of a large set of non-homologous 3ß-corner structures of various origins. We focused on geometric parameters such as change in gyration radius, solvent-accessible area, major conformer lifetime and torsion angles, and the number of hydrogen bonds. Ultimately, a set of 3ß-corners from 330 structures was characterized by a root mean square deviation (RMSD) of less than 5 Å, a change in the gyration radius of no more than 5%, and the preservation of amino acid residues positioned within the allowed regions on the Ramachandran map. The studied structures retained their topologies throughout the MD experiments. Thus, the 3ß-corner structure was found to be rather stable per se in a water environment, i.e., without the rest of a protein molecule, and can act as the nucleus or "ready-made" building block in protein folding. The 3ß-corner can also be considered as an independent object for study in field of structural biology.

In Silico Study of the Interactions of Anle138b Isomer, an Inhibitor of Amyloid Aggregation, with Partner Proteins.

Herein, we aimed to highlight current "gaps" in the understanding of the potential interactions between the Anle138b isomer ligand, a promising agent for clinical research, and the intrinsically disordered alpha-synuclein protein. The presence of extensive unstructured areas in alpha-synuclein determines its existence in the cell of partner proteins, including the cyclophilin A chaperone, which prevents the aggregation of alpha-synuclein molecules that are destructive to cell life. Using flexible and cascaded molecular docking techniques, we aimed to expand our understanding of the molecular architecture of the protein complex between alpha-synuclein, cyclophilin A and the Anle138b isomer ligand. We demonstrated the possibility of intricate complex formation under cellular conditions and revealed that the main interactions that stabilize the complex are hydrophobic and involve hydrogen.

Atomic Simulation of the Binding of JAK1 and JAK2 with the Selective Inhibitor Ruxolitinib.

Rheumatoid arthritis belongs to the group of chronic systemic autoimmune diseases characterized by the development of destructive synovitis and extra-articular manifestations. Cytokines regulate a wide range of inflammatory processes involved in the pathogenesis of rheumatoid arthritis and contribute to the induction of autoimmunity and chronic inflammation. Janus-associated kinase (JAK) and signal transducer and activator of transcription (STAT) proteins mediate cell signaling from cytokine receptors, and are involved in the pathogenesis of autoimmune and inflammatory diseases. Targeted small-molecule drugs that inhibit the functional activity of JAK proteins are used in clinical practice for the treatment of rheumatoid arthritis. In our study, we modeled the interactions of the small-molecule drug ruxolitinib with JAK1 and JAK2 isoforms and determined the binding selectivity using molecular docking. Molecular modeling data show that ruxolitinib selectively binds the JAK1 and JAK2 isoforms with a binding affinity of -8.3 and -8.0 kcal/mol, respectively. The stabilization of ligands in the cavity of kinases occurs primarily through hydrophobic interactions. The amino acid residues of the protein globules of kinases that are responsible for the correct positioning of the drug ruxolitinib and its retention have been determined.

Convolutional neural network in proteomics and metabolomics for determination of comorbidity between cancer and schizophrenia.

The association between cancer risk and schizophrenia is widely debated. Despite many epidemiological studies, there is still no strong evidence regarding the molecular basis for the comorbidity between these two pathological conditions. The vast majority of assays have been performed using clinical records of schizophrenic patients or those undergoing cancer treatment and monitored for sufficient time to find shared features between the considered conditions. We performed mass spectrometry-based proteomic and metabolomic investigations of patients with different cancer phenotypes (breast, ovarian, renal, and prostate) and patients with schizophrenia. The resulting vast quantity of proteomic and metabolomic data were then processed using systems biology and one-dimensional (1D) convolutional neural network (1DCNN) machine learning approaches. Traditional systematic approaches permit the segregation of schizophrenia and cancer phenotypes on the level of biological processes, while 1DCNN recognized "signatures" that could segregate distinct cancer phenotypes and schizophrenia at the comorbidity level. The designed network efficiently discriminated unrelated pathologies with a model accuracy of 0.90 and different subtypes of oncophenotypes with an accuracy of 0.94. The proposed strategy integrates systematic analysis of identified compounds and application of 1DCNN model for unidentified ones to reveal the similarity between distinct phenotypes.

Tracking Health, Performance and Recovery in Athletes Using Machine Learning.

Training and competitive periods can temporarily impair the performance of an athlete. This disruption can be short- or long-term, lasting up to several days. We analyzed the health indicators of 3661 athletes during an in-depth medical examination. At the time of inclusion in the study, the athletes were healthy. Instrumental examinations (fluorography, ultrasound examination of the abdominal cavity and pelvic organs, echocardiography, electrocardiography, and stress testing "to failure"), laboratory examinations (general urinalysis and biochemical and general clinical blood analysis), and examinations by specialists (ophthalmologist, otolaryngologist, surgeon, cardiologist, neurologist, dentist, gynecologist (women), endocrinologist, and therapist) were performed. This study analyzed the significance of determining the indicators involved in the implementation of the "catabolism" and "anabolism" phenotypes using the random forest and multinomial logistic regression machine learning methods. The use of decision forest and multinomial regression models made it possible to identify the most significant indicators of blood and urine biochemistry for the analysis of phenotypes as a characterization of the effectiveness of recovery processes in the post-competitive period in athletes. We found that the parameters of muscle metabolism, such as aspartate aminotransferase, creatine kinase, lactate dehydrogenase, and alanine aminotransferase levels, and the parameters of the ornithine cycle, such as creatinine, urea acid, and urea levels, made the most significant contribution to the classification of two types of metabolism: catabolism and anabolism.

Managing of Unassigned Mass Spectrometric Data by Neural Network for Cancer Phenotypes Classification.

Mass spectrometric profiling provides information on the protein and metabolic composition of biological samples. However, the weak efficiency of computational algorithms in correlating tandem spectra to molecular components (proteins and metabolites) dramatically limits the use of "omics" profiling for the classification of nosologies. The development of machine learning methods for the intelligent analysis of raw mass spectrometric (HPLC-MS/MS) measurements without involving the stages of preprocessing and data identification seems promising. In our study, we tested the application of neural networks of two types, a 1D residual convolutional neural network (CNN) and a 3D CNN, for the classification of three cancers by analyzing metabolomic-proteomic HPLC-MS/MS data. In this work, we showed that both neural networks could classify the phenotypes of gender-mixed oncology, kidney cancer, gender-specific oncology, ovarian cancer, and the phenotype of a healthy person by analyzing 'omics' data in 'mgf' data format. The created models effectively recognized oncopathologies with a model accuracy of 0.95. Information was obtained on the remoteness of the studied phenotypes. The closest in the experiment were ovarian cancer, kidney cancer, and prostate cancer/kidney cancer. In contrast, the healthy phenotype was the most distant from cancer phenotypes and ovarian and prostate cancers. The neural network makes it possible to not only classify the studied phenotypes, but also to determine their similarity (distance matrix), thus overcoming algorithmic barriers in identifying HPLC-MS/MS spectra. Neural networks are versatile and can be applied to standard experimental data formats obtained using different analytical platforms.