Search

The article discusses a new method called ImmuneCLIP that aims to improve the understanding of T-cell receptor (TCR) and epitope interactions in the human immune system. ImmuneCLIP utilizes pre-trained protein language models to align TCR and epitope embeddings in a shared latent space, outperforming traditional sequence-similarity based methods and existing deep learning models in epitope ranking and binding prediction tasks. The method shows promise in studying diverse immune interactions and enhancing our comprehension of human immune recognition systems, even with limited training data. [Extracted from the article]

A study from the Windland Smith Rice Sudden Death Genomics Laboratory in Rochester, Minnesota, explores the use of deep protein language models in resolving variants of uncertain significance (VUS) in genes related to long-QT syndrome. The research compares the performance of these models with traditional patch-clamp techniques and finds that while the models show high sensitivity, they have lower specificity. The study suggests that while these tools can aid in reducing the number of variants needing further analysis, they cannot fully replace functional characterization. [Extracted from the article]

A recent study conducted at Universiti Kebangsaan Malaysia focused on the effectiveness of lyoprotectants in protein stabilization during lyophilization, a process commonly used in the healthcare industry. The research highlighted the importance of choosing suitable lyoprotectants to protect protein conformations and increase shelf life. The study concluded that disaccharides were the most effective lyoprotectants, while other groups were best used in combination with other lyoprotectants. This research provides valuable insights for the pharmaceutical industry and healthcare professionals. [Extracted from the article]

The article discusses the challenges in distinguishing cancer-causing driver mutations from random passenger mutations and introduces the AI-based D2Deep method to address this issue. D2Deep combines a non-specialized protein language model with protein-specific evolutionary information to outperform existing predictors and offer result interpretation. The model, trained on a balanced somatic training set, helps mitigate biases related to hotspot mutations and can be used for mutation prioritization in clinical settings. The preprint emphasizes the importance of unbiased predictive models in cancer research and provides a link for accessing D2Deep predictions and confidence scores for clinical interpretation. [Extracted from the article]

The article discusses a study that aims to predict and design druggable proteins by analyzing FDA-approved proteins and developing machine learning models. The researchers used a dataset of FDA-approved proteins and randomly selected proteins for training and validation, achieving an AUC of 0.80 with MCC 0.61 on validation data. They also performed motif analysis and developed an ensemble-based method with a web server and standalone package to aid in predicting and designing druggable proteins. The study has not been peer-reviewed and provides valuable insights into the field of bioinformatics and biotechnology. [Extracted from the article]

Researchers from Shandong University have developed a new neural network called PepNet that can accurately predict anti-inflammatory peptides (AIPs) and antimicrobial peptides (AMPs). AIPs and AMPs are important for developing peptide-based therapies for inflammation and microbial infections. PepNet utilizes a pre-trained protein language model to fully utilize peptide sequence information, and it outperforms other predictors in terms of accuracy and interpretability. The researchers have also made a user-friendly web server for PepNet freely available. This study was published in Communications Biology. [Extracted from the article]

A preprint abstract from biorxiv.org discusses a new method called pKALM for predicting protein pKa values. The method uses a protein language model (PLM) to capture the sequence-structure relationship of proteins and achieves state-of-the-art performance. pKALM accurately predicts the pKa values of six residues and two termini with high precision and efficiency, including both exposed and buried residues. The study also suggests that predicted protein isoelectric points (pI) can improve the accuracy of pKa prediction. The method is expected to be a valuable tool for researchers in biochemistry, biophysics, and drug design. [Extracted from the article]

A recent preprint abstract from biorxiv.org discusses the maturation pathway of a eukaryotic virus, specifically the non-enveloped, T=4, ssRNA-containing Nudaurelia capensis omega virus (N{omega}V). The study used cryo-electron microscopy (cryo-EM) to analyze the structures of five intermediates in the maturation process of the virus. The researchers were able to observe unique trajectories and refolding of subunit regions, including the formation of an autocatalytic cleavage site. This study provides a detailed description of the maturation pathway for a eukaryotic virus, which has been elusive in previous research. However, it is important to note that this preprint has not yet undergone peer review. [Extracted from the article]

A recent study conducted by researchers at Guangxi University in China explored the impact of estrogen levels on voiding characteristics and bladder structure in a mouse model of menopause-induced lower urinary tract symptoms (LUTS). The study found that lower estrogen levels were associated with increased voiding episodes and structural changes in the bladder tissues, such as increased collagen deposition. Additionally, the study identified a negative relationship between estrogen levels and the expression of mechanosensitive proteins in the bladder tissues, which may influence voiding frequency and nocturia. The findings suggest that estrogen treatment could potentially improve bladder function and reduce urinary frequency in postmenopausal women. [Extracted from the article]

A recent preprint abstract from biorxiv.org discusses the development of a potential therapeutic candidate for toxic shock syndrome caused by the bacteria Paeniclostridium sordellii. This disease primarily affects postpartum and post-abortive women and has a high mortality rate. The researchers used a protein diffusion model to design a potent inhibitor for the lethal toxin produced by P. sordellii. The inhibitor showed promising results in protecting mice from lung edema and death in a lethal challenge model. This research demonstrates the potential of deep learning-based protein design in rapidly generating biologics with clinical utility. However, it is important to note that this preprint has not yet undergone peer review. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the development of a method called proseLM for protein sequence design. This method adapts protein language models to incorporate structural and functional context, improving the recovery of native residues during design. The addition of non-protein context, such as nucleic acids, ligands, and ions, further enhances the recovery of native residues. The effectiveness of proseLM was experimentally validated by optimizing the editing efficiency of genome editors in human cells and redesigning therapeutic antibodies. However, it is important to note that this preprint has not yet undergone peer review. [Extracted from the article]

A recent research paper from King Faisal University in Saudi Arabia explores the use of bimodal skew-symmetric distributions in analyzing cancer cell protein data. These distributions are characterized by parameters that control asymmetry and kurtosis. The study found that the proposed model, which incorporates these distributions, outperformed four established distributions in terms of analytical precision. This research has the potential to enhance biological research in the field of cancer. For more information, the full article can be accessed for free through the provided link. [Extracted from the article]

The article focuses on the utilization of an optimized AlphaFold protein model for structure-based design of a selective HDAC11 inhibitor with anti-neuroblastoma activity, highlighting the potential of this approach for drug discovery. Topics include the optimization of the AlphaFold model for correct prediction of folding in the absence of ligands and cofactors, the design of a new scaffold for HDAC11 inhibitors.

A recent study conducted by researchers at Kennesaw State University in Marietta, Georgia, explores the use of artificial intelligence (AI) in protein structure prediction. The study provides a comprehensive overview of computational models used in predicting protein structure, including established protein modeling techniques and deep learning-based models. The researchers discuss the integration of AI techniques into established frameworks and compare the performance of different models. The study also highlights the applications of protein structure prediction in various fields such as drug design, industry, education, and novel protein development. Overall, the research emphasizes the significant advancements achieved by AI and identifies areas for further investigation. [Extracted from the article]

A new study conducted by researchers at Zhejiang University in China has developed a highly accurate and efficient deep learning paradigm called KarmaLoop for protein loop modeling. Protein loop modeling is a challenging task in protein structure prediction, and existing methods have limitations in terms of accuracy and computational efficiency. KarmaLoop outperforms conventional and deep learning-based methods in terms of accuracy and efficiency, with potential applications in protein engineering, antibody-antigen recognition, and drug design. This research has the potential to advance the field of science and technology. [Extracted from the article]

A preprint abstract discusses the development of a machine learning model for predicting plasma protein binding (PPB) of drugs. PPB is important for understanding drug behavior in the body. The model was validated and found to have high accuracy and interpretability. It outperformed other existing models in predicting PPB for different compounds. The study also identified structural features related to PPB. The model is publicly available on the OCHEM platform. This research has not yet undergone peer review. [Extracted from the article]

A recent study conducted by researchers at the University of Texas Rio Grande Valley explores the spatiotemporal dynamics of the Unfolded Protein Response (UPR) mechanism in prion diseases. The UPR is a cellular stress response that can slow or stop protein translation and reduce toxic protein levels. However, it can also trigger processes leading to cell death and is implicated in the progression of neurodegenerative diseases. The researchers developed a mathematical model centered around a single neuron to describe the UPR mechanism and found that certain modifications, such as decreasing the production of healthy proteins and increasing the clearance rate of toxic proteins, may help reduce disease progression. This research provides insights into potential drug therapies for prion diseases. [Extracted from the article]

A preprint abstract from medrxiv.org discusses the use of DNA methylation (DNAm) models to understand non-genetic variation in protein abundance across the lifespan. The study evaluated 14 previously published DNAm models in a large adult population and found that the models performed better at older ages than in children. The findings suggest that DNAm models can capture non-genetic variation in protein abundance and provide valuable insights into exposure, phenotype, and disease risk. However, it is important to note that this preprint has not yet undergone peer review. [Extracted from the article]

Researchers from the University of New Orleans have developed a computational method called DisPredict3.0 to predict intrinsically disordered regions (IDRs) in proteins. IDRs are protein regions that do not have a stable three-dimensional structure and can contribute to various diseases. The method utilizes deep learning techniques and evolutionary information from protein language models to improve the accuracy of IDR prediction. DisPredict3.0 has shown promising results and outperformed existing methods in terms of various evaluation metrics. The research has been peer-reviewed and the DisPredict3.0 webserver is available for use. [Extracted from the article]

A recent study conducted by researchers at Technical University Dresden (TU Dresden) in Germany explored the molecular alterations in skeletal muscle (SKM) in heart failure with preserved ejection fraction (HFpEF). The study focused on the modulation of titin, titin phosphorylation, and contraction-regulating proteins in limb SKM and the diaphragm. The results showed hyperphosphorylation of titin in limb SKM, which could be reversed by high-intensity interval training (HIIT). Additionally, the study found significant differences in the expression of contraction-regulating proteins between the diaphragm and limb SKM, suggesting muscle atrophy and dysfunction in limb SKM but not in the diaphragm. These findings indicate a correlation between titin stiffening and exercise intolerance in HFpEF, as well as differential regulation between different SKM groups. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the limitations of current protein language models, which primarily operate at the residue scale and do not provide information at the atom level. The authors propose a novel approach called ESM-AA (ESM All-Atom) that enables atom-scale and residue-scale unified molecular modeling. Experimental results show that ESM-AA outperforms previous methods in protein-molecule tasks, demonstrating the full utilization of protein language models. The source codes for ESM-AA are publicly available. However, it is important to note that this preprint has not been peer-reviewed. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the limitations of current protein language models in providing information at the atom level, which hinders their application in protein engineering. The authors propose a novel approach called ESM-AA (ESM All-Atom) that enables atom-scale and residue-scale unified molecular modeling. By pre-training on multi-scale code-switch protein sequences and utilizing a multi-scale position encoding, ESM-AA surpasses previous methods in protein-molecule tasks and retains its understanding of proteins. This research has not yet undergone peer review. [Extracted from the article]

Researchers at Zhengzhou University of Light Industry in China have developed a new method for accurately identifying antimicrobial peptides (AMPs) using deep learning networks and pre-trained protein language models. The method, called UniproLcad, integrates three prominent protein language models to extract multi-view features from AMP sequences and assign attention weights to them. Through cross-validation and independent testing, UniproLcad demonstrated competitive performance in identifying AMPs. This research contributes to the advancement of computational methods in the field of antimicrobial peptide identification. [Extracted from the article]

A recent study conducted by researchers at the University of Notre Dame focused on identifying protein biomarkers that differentiate neutrophils and myeloid-derived suppressor cells (MDSCs). MDSCs are known to suppress the activity of T lymphocytes in solid tumors, making them an attractive target for immunotherapy. The researchers used a human isogenic cell line model and protein profiling to analyze the protein expression patterns of these cells. They identified a 38-protein signature that is upregulated in MDSCs compared to normal monocytes or neutrophils. These findings could potentially lead to the development of therapeutic targets that specifically impair MDSCs without harming healthy myeloid cells. [Extracted from the article]

A recent study published in Health & Medicine Week discusses the importance of accurately predicting protein solubility in various biotechnological, industrial, and biomedical applications. The study focuses on the use of attention-based protein language models (PLMs) to enhance the prediction of protein solubility. The researchers curated an updated dataset and developed a model called PLM_Sol, which demonstrated significant improvements in accuracy, F1_score, and MCC score compared to previous models. The study suggests that PLM_Sol could be valuable for large-scale studies on the development of novel functional enzymes. However, it is important to note that this study has not yet undergone peer review. A recent study published in Health & Medicine Week discusses the importance of accurately predicting protein solubility in various biotechnological, industrial, and biomedical applications. The study focuses on the use of attention-based protein language models (PLMs) to enhance the prediction of protein solubility. The researchers curated an updated dataset and developed a model called PLM_Sol, which demonstrated significant improvements in accuracy, F1_score, and MCC score compared to previous models. The study suggests that PLM_Sol could be valuable for large-scale studies on the development of novel functional enzymes. However, it is important to note that this study has not yet undergone peer review. [Extracted from the article]

According to a preprint abstract from biorxiv.org, researchers have developed a new method for virtual screening in drug design that does not require high-resolution three-dimensional protein structures. Instead, they use protein language models and molecular graphs as inputs to a graph-to-transformer cross-attention mechanism. This approach achieves a screening power comparable to traditional structure-based models, but with greatly reduced computational requirements. The implications of this research include faster virtual screening and the ability to perform early stages of computer-aided drug design without 3D protein structures. Please note that this preprint has not yet undergone peer review. [Extracted from the article]

A recent study conducted by researchers at the Massachusetts Institute of Technology (MIT) has found that generative artificial intelligence (AI) is revolutionizing computational structural biology. The study highlights the use of diffusion models (DM) in protein structure generation and small molecule docking, which have achieved state-of-the-art results. The researchers provide an overview of DMs, their applications in structural biology, and the challenges and potential of generative techniques in this field. The study was supported by the Abdul Latif Jameel Clinic for Machine Learning in Health at MIT and has been peer-reviewed. [Extracted from the article]

A study conducted at the Federal University Sao Paulo in Brazil examined the effects of photobiomodulation on synovial morphology, gene expression, and protein expression in a model of acute inflammation. The study used female Wistar rats divided into three groups: control, sham (arthritis), and PBM (arthritis and photobiomodulation). The results showed that photobiomodulation reduced inflammation, upregulated iNOS gene expression, and downregulated protein expression compared to the sham group. This research provides insights into potential treatments for arthritis and highlights the benefits of photobiomodulation therapy. [Extracted from the article]

According to a preprint abstract, researchers have found that the accumulation of Aβ protein in the brain, a hallmark of Alzheimer's disease, leads to dysfunction in the nuclear pore complex (NPC), which regulates the movement of molecules in and out of the cell nucleus. This dysfunction results in the degradation of the permeability barrier and inefficient transport of proteins. The study suggests that this NPC dysfunction makes neurons more susceptible to inflammation-induced necroptosis, a programmed cell death pathway. The findings highlight the potential of targeting the NPC as a therapeutic approach for neurodegenerative diseases. However, it is important to note that this research has not yet undergone peer review. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the development of a new model called ProtAlign-ARG, which combines a pre-trained protein language model with an alignment scoring-based model to identify and classify antibiotic resistance genes (ARGs). The model utilizes raw protein language model embeddings to classify ARGs and incorporates an alignment-based scoring method when lacking confidence. ProtAlign-ARG demonstrates high accuracy in identifying and classifying ARGs, particularly excelling in recall compared to existing tools. The preprint has not yet undergone peer review. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the limitations of current protein language models, which primarily operate at the residue scale and do not provide information at the atom level. The abstract introduces a novel approach called ms-ESM (multi-scale ESM) that enables multi-scale unified molecular modeling by pre-training on multi-scale code-switch protein sequences and utilizing a multi-scale position encoding. Experimental results show that ms-ESM outperforms previous methods in protein-molecule tasks, demonstrating the full utilization of protein language models. The abstract also mentions that ms-ESM gains molecular knowledge while retaining its understanding of proteins through unified molecular modeling. However, it is important to note that this preprint has not been peer-reviewed. [Extracted from the article]

A recent study published in Drug Week discusses the challenges in accurately predicting the structures of protein complexes involving phosphopeptides, which are peptides that have been phosphorylated. The study addresses these limitations by refining a computational biology tool called AlphaFold to improve its accuracy in modeling these complex structures. The enhanced model outperforms the original AlphaFold, showing a substantial increase in accuracy scores. This research has implications for understanding cellular signaling and could contribute to the development of innovative therapies. It is important to note that this study has not yet undergone peer review. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the development of a machine learning-assisted antibody generation pipeline that accelerates the screening and re-design of immunoglobulins G (IgGs) against a broad spectrum of coronavirus variants. The pipeline uses protein structural docking models to assess the interaction between IgGs and the viral spike's receptor binding domain (RBD) and predicts their viral neutralizing activities. Additionally, Gaussian process regression and antibody language models are employed to predict the activity profiles of IgGs against viral strains. The research aims to expedite the application of artificial intelligence in low-data regimes for antibody analysis and efficacy improvement. However, it is important to note that this preprint has not been peer-reviewed. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the development of a new approach called ProT-Diff for generating antimicrobial peptide (AMP) sequences. AMPs are important for combating antibiotic-resistant pathogens, but current methods for discovering new AMPs have limitations. ProT-Diff combines a pre-trained protein language model with a diffusion model to efficiently generate thousands of AMPs with diverse lengths. In silico screening and experimental validation identified 34 AMPs with antimicrobial activity, including one with broad-spectrum efficacy. This study presents a promising strategy for designing novel AMPs and has potential applications in new drug development. [Extracted from the article]

A preprint abstract from biorxiv.org discusses the potential of controllable protein design using protein language models (ProtLMs). These models are inspired by the success of pre-trained language models in natural language processing tasks and aim to generate protein sequences with specific functions. The study proposes a method called PrefixProt, which utilizes prefix tuning to learn virtual tokens as control tags for protein generation. The results show that prefix virtual tokens can efficiently prompt ProtLMs, especially in low-data settings, and combining virtual tokens improves the proportion of generated proteins with multiple properties. The authors believe that PrefixProt could contribute to drug discovery and biomedical advancement. However, it's important to note that this preprint has not been peer-reviewed. [Extracted from the article]

A recent study conducted at Massachusetts General Hospital has identified serum proteins that can distinguish between Crohn's disease (CD) and ulcerative colitis (UC) in pediatric patients. The researchers used the SomaScan proteomics platform to measure 11,000 proteins in the serum of CD and UC patients. They found that a 4-protein model showed high accuracy in discriminating between CD and UC, with an area under the curve (AUC) of 0.97. This research not only provides potential diagnostic biomarkers but also sheds light on the immune and metabolic pathways that differentiate CD from UC. [Extracted from the article]

According to a preprint abstract, researchers have found that the accumulation of Aβ, a protein associated with Alzheimer's disease, leads to dysfunction in the nuclear pore complex (NPC) in neurons. The NPC is responsible for regulating the movement of macromolecules across the nuclear envelope, and its impairment has been observed in various neurodegenerative disorders. The dysfunction of the NPC in Alzheimer's disease neurons makes them more susceptible to inflammation-induced necroptosis, a programmed cell death pathway. This research suggests that the NPC could be a potential therapeutic target for Alzheimer's disease. However, it is important to note that this preprint has not yet undergone peer review. [Extracted from the article]

A recent study conducted at Kansas State University has developed a protein language model (pLM) with evolutionary scale modeling (ESM-2) embeddings to screen peptides with strong ACE inhibitory activity. The researchers tested various embedding approaches and machine learning modeling methods, finding that logistic regression with ESM-2 embeddings showed the best performance. The study concluded by making the user-friendly webserver with the top three models freely available. This research provides valuable insights into the development of antihypertensive peptides and offers potential applications in clinical treatment for hypertension. [Extracted from the article]

A recent research study conducted by the Faculty of Mathematics at Baotou Teachers' College in China explores the oscillatory behavior of the p53 protein, which plays a crucial role in the body's defense against cancer. The study investigates the impact of time delays on p53 oscillations and suggests that discrete time delays promote these oscillations, while distributed time delays have a dual effect. The findings indicate that controlling the time delay within an appropriate range may enhance the inhibitory effect of p53 on cancer cells. This research provides valuable insights into the mechanisms underlying p53 oscillations and their potential implications for cancer treatment. [Extracted from the article]

Researchers at the Mayo Clinic in Rochester, Minnesota have conducted a study on the prediction of variant pathogenicity in the low density lipoprotein receptor (LDLR) protein structure. They compared the effectiveness of deep generative models, including Evolutionary model of Variant Effect (EVE), Evolutionary Scale Modeling (ESM), and AlphaFold 2 (AF2), with established methods for predicting variant pathogenicity. The study found that AF2 poorly performed in these tasks, while ESM and EVE were competitive with prior scoring methods and showed stronger correlations with experimental assays and clinical phenotypes. The research suggests that AF2 predicted LDLR structures do not accurately model variant pathogenicity. [Extracted from the article]

A recent report from Michigan State University discusses the use of transgenic amyloid precursor protein (APP) mouse models in studying Alzheimer's disease (AD). These models accurately mimic certain aspects of AD, such as amyloid deposits, and have been used to predict the effectiveness of anti-amyloid therapies. The report highlights the success of immunotherapy in reducing amyloid and cognitive impairment in mice, as well as the predictive value of these early APP models in designing therapeutic treatments. While new mouse models may further enhance research on AD, the early models have proven valuable in understanding amyloid therapies in humans. [Extracted from the article]

The article discusses the importance of predicting the effects of mutations in proteins for various applications, such as understanding genetic diseases and designing proteins to address challenges in climate, agriculture, and healthcare. It highlights the challenges in assessing the benefits of machine learning-based protein models due to the use of different experimental datasets and variable performance across protein families. To address these challenges, the authors introduce ProteinGym, a large-scale benchmark that includes standardized deep mutational scanning assays and curated clinical datasets. They also provide a robust evaluation framework and open-source the corresponding codebase, datasets, and model predictions. The article emphasizes that the preprint has not been peer-reviewed. [Extracted from the article]

A research study conducted in Madhya Pradesh, India, focused on in silico mutational research on the energetics of the HIV-1 NNRTIs protein. The study generated 380 unique proteins in silico and analyzed their impact on folding-unfolding free energy, protein stability, and solvation energy. The findings showed that out of the 380 proteins, 11 had extremely unfavorable folding-unfolding changes, 69 had less-favorable changes, and 270 had favorable changes. This research has been peer-reviewed and provides valuable insights into the stability of designed proteins. [Extracted from the article]

A recent study conducted in Odisha, India, focused on the development of pharmacophore models for the AcrB protein and the identification of potential adjuvant candidates for overcoming efflux-mediated colistin resistance. The researchers aimed to find an FDA-approved efflux inhibitor that could be used in combination with colistin to enhance its effectiveness. Through molecular docking and in vitro evaluation, they identified argatroban as a top non-antibiotic hit that significantly inhibited the efflux activity of colistin-resistant clinical isolates. The combination of argatroban and colistin showed synergistic effects in reducing bacterial growth. This study provides valuable insights into potential strategies for combating multi-drug resistance in ESKAPE pathogens. [Extracted from the article]

A preprint abstract from biorxiv.org discusses a new method called PrefixProt for designing proteins with specific structures and functions. The method uses prefix-tuning to learn virtual tokens for different protein properties, which can then be used to prompt pre-trained protein language models (ProtLMs) to generate tailored proteins. The study demonstrates that prefix virtual tokens are effective in controlling protein generation, even with limited data. The authors believe that PrefixProt has the potential to contribute to protein discovery and biomedical advancements. However, it is important to note that this preprint has not yet undergone peer review. [Extracted from the article]

Researchers at the University of Rome Tor Vergata have conducted a study on the COL2A1 gene, which encodes the alpha-1 chain of type-II procollagen. They aimed to characterize a specific variant of this gene, known as p.Gly444Ser, which was detected in a fetus with a severe skeletal phenotype and in the mild-affected father. Through functional studies, protein modeling, and clinical assessment, the researchers found that this variant alters the processing and assembly of collagen-II molecules, leading to a mild form of type-II collagenopathy. This study demonstrates the importance of combining genomics and clinical expertise to establish the pathogenicity of genetic variants. [Extracted from the article]

Researchers from Instituto Politecnico Nacional in Mexico City have proposed a consolidated structural model of the CagY protein of Helicobacter pylori (Hp), a bacterium associated with gastric inflammation and gastric cancer. CagY is a large and complex protein that plays a critical role in Hp's type IV secretion system (T4SS), which injects Hp molecules into human gastric cells. The researchers used various modeling techniques to construct a 1595 amino acid chain and a 14-chain CagY multimer structure, which revealed the formation of the T4SS channel and the potential function of the middle repeat region (MRR) in modulating tissue inflammation. This study provides valuable insights into the structure and function of CagY and its role in Hp infection. [Extracted from the article]

A recent study conducted by researchers at Islamic Azad University in Marvdasht, Iran, has identified a novel missense variant in the CDK5RAP2 gene that is associated with non-obstructive azoospermia (NOA), a severe form of male infertility. The researchers used exome-sequencing, protein modeling, and gene expression data analysis to investigate a consanguineous family with three NOA brothers and one fertile brother. They found that the three NOA brothers were homozygous for a rare missense variant in the CDK5RAP2 gene, which may lead to defective centrosomic maturation during spermatogenesis. This research provides new insights into the genetic basis of NOA and male infertility. [Extracted from the article]

A recent preprint abstract discusses the development of a new model called FrameDiPT for protein structure inpainting. This model is important for T-cells due to the hyper-variability of the complementarity determining region (CDR) loops. The researchers evaluated the model on CDR loop design for T-cell receptors and achieved comparable prediction accuracy to other existing models. FrameDiPT is a generative model that captures the conformational distribution at different regions and binding states, providing an advantage over deterministic structure prediction models. However, it is important to note that this preprint has not yet undergone peer review. [Extracted from the article]