Search

Motivation Protein structure determination has primarily been performed using X-ray crystallography. To overcome the expensive cost, high attrition rate and series of trial-and-error settings, many in-silico methods have been developed to predict crystallization propensities of proteins based on their sequences. However, the majority of these methods build their predictors by extracting features from protein sequences, which is computationally expensive and can explode the feature space. We propose DeepCrystal, a deep learning framework for sequence-based protein crystallization prediction. It uses deep learning to identify proteins which can produce diffraction-quality crystals without the need to manually engineer additional biochemical and structural features from sequence. Our model is based on convolutional neural networks, which can exploit frequently occurring k -mers and sets of k -mers from the protein sequences to distinguish proteins that will result in diffraction-quality crystals from those that will not. Results Our model surpasses previous sequence-based protein crystallization predictors in terms of recall, F -score, accuracy and Matthew's correlation coefficient (MCC) on three independent test sets. DeepCrystal achieves an average improvement of 1.4, 12.1% in recall, when compared to its closest competitors, Crysalis II and Crysf, respectively. In addition, DeepCrystal attains an average improvement of 2.1, 6.0% for F-score, 1.9, 3.9% for accuracy and 3.8, 7.0% for MCC w.r.t. Crysalis II and Crysf on independent test sets. Availability and implementation The standalone source code and models are available at https://github.com/elbasir/DeepCrystal and a web-server is also available at https://deeplearning-protein.qcri.org. Supplementary information Supplementary data are available at Bioinformatics online. [ABSTRACT FROM AUTHOR]

Motivation: Protein solubility can be a decisive factor in both research and production efficiency and in silico sequence-based predictors that can accurately estimate solubility outcomes are highly sought. Results: In this study, we present a novel approach termed PRotein SolubIlity Predictor (PaRSnIP), which uses a gradient boosting machine algorithm as well as an approximation of sequence and structural features of the protein of interest. Based on an independent test set, PaRSnIP outperformed other state-of-the-art sequence-based methods by more than 9% in accuracy and 0.17 in Matthew's correlation coefficient, with an overall accuracy of 74% and Matthew's correlation coefficient of 0.48. Additionally, PaRSnIP provides importance scores for all features used in training. We observed higher fractions of exposed residues to associate positively with protein solubility and tripeptide stretches with multiple histidines to associate negatively with solubility. The improved prediction accuracy of PaRSnIP should enable it to predict protein solubility with greater reliability and to screen for sequence variants with enhanced manufacturability. [ABSTRACT FROM AUTHOR]

The COVID-19 pandemic, caused by the β-coronavirus (β-CoV) severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), continues to cause significant global morbidity and mortality. While vaccines have reduced the overall number of severe infections, there remains an incomplete understanding of viral entry and innate immune activation, which can drive pathology. Innate immune responses characterized by positive feedback between cell death and cytokine release can amplify the inflammatory cytokine storm during β-CoV–mediated infection to drive pathology. Therefore, there remains an unmet need to understand innate immune processes in response to β-CoV infections to identify therapeutic strategies. To address this gap, here we used an MHV model and developed a whole genome CRISPR-Cas9 screening approach to elucidate host molecules required for β-CoV infection and inflammatory cell death, PANoptosis, in macrophages, a sentinel innate immune cell. Our screen was validated through the identification of the known MHV receptor Ceacam1 as the top hit, and its deletion significantly reduced viral replication due to loss of viral entry, resulting in a downstream reduction in MHV-induced cell death. Moreover, this screen identified several other host factors required for MHV infection-induced macrophage cell death. Overall, these findings demonstrate the feasibility and power of using genome-wide PANoptosis screens in macrophage cell lines to accelerate the discovery of key host factors in innate immune processes and suggest new targets for therapeutic development to prevent β-CoV-induced pathology. A CRISPR screen identified host factors that promote viral entry and inflammatory cell death, PANoptosis, upon infection with the murine hepatitis virus, a type of beta-coronavirus, suggesting potential therapeutic targets for coronavirus infections. [ABSTRACT FROM AUTHOR]

Background: While high-grade serous ovarian cancer (HGSC) has proven largely resistant to immunotherapy, sporadic incidents of partial and complete response have been observed in clinical trials and case reports. These observations suggest that a molecular basis for effective immunity may exist within a subpopulation of HGSC. Herein, we developed an algorithm, CONSTRU (Computing Prognostic Marker Dependencies by Successive Testing of Gene-Stratified Subgroups), to facilitate the discovery and characterization of molecular backgrounds of HGSC that confer resistance or susceptibility to protective anti-tumor immunity. Methods: We used CONSTRU to identify genes from tumor expression profiles that influence the prognostic power of an established immune cytolytic activity signature (CYTscore). From the identified genes, we developed a stratification signature (STRATsig) that partitioned patient populations into tertiles that varied markedly by CYTscore prognostic power. The tertile groups were then analyzed for distinguishing biological, clinical and immunological properties using integrative bioinformatics approaches. Results: Patient survival and molecular measures of immune suppression, evasion and dysfunction varied significantly across STRATsig tertiles in validation cohorts. Tumors comprising STRATsig tertile 1 (S-T1) showed no immune-survival benefit and displayed a hyper-immune suppressed state marked by activation of TGF-β, Wnt/β-catenin and adenosine-mediated immunosuppressive pathways, with concurrent T cell dysfunction, reduced potential for antigen presentation, and enrichment of cancer-associated fibroblasts. By contrast, S-T3 tumors exhibited diminished immunosuppressive signaling, heightened antigen presentation machinery, lowered T cell dysfunction, and a significant CYTscore-survival benefit that correlated with mutational burden in a manner consistent with anti-tumor immunoediting. These tumors also showed elevated activity of DNA damage/repair, cell cycle/proliferation and oxidative phosphorylation, and displayed greater proportions of Th1 CD4 + T cells. In these patients, but not those of S-T1 or S-T2, validated predictors of immunotherapy response were prognostic of longer patient survival. Further analyses showed that STRATsig tertile properties were not explained by known HGSC molecular or clinical subtypes or singular immune mechanisms. Conclusions: STRATsig is a composite of parallel immunoregulatory pathways that mirrors tumor immunogenic potential. Approximately one-third of HGSC cases classify as S-T3 and display a hypo-immunosuppressed and antigenic molecular composition that favors immunologic tumor control. These patients may show heightened responsiveness to current immunotherapies. [ABSTRACT FROM AUTHOR]

Objective: To develop a machine-based algorithm from clinical and demographic data, physical activity and glucose variability to predict hyperglycaemic and hypoglycaemic excursions in patients with type 2 diabetes on multiple glucose lowering therapies who fast during Ramadan.Patients and Methods: Thirteen patients (10 males and three females) with type 2 diabetes on 3 or more anti-diabetic medications were studied with a Fitbit-2 pedometer device and Freestyle Libre (Abbott Diagnostics) 2 weeks before and 2 weeks during Ramadan. Several machine learning techniques were trained to predict blood glucose levels in a regression framework utilising physical activity and contemporaneous blood glucose levels, comparing Ramadan to non-Ramadan days.Results: The median age of participants was 51 years (IQR 49-52); median BMI was 33.2 kg/m2 (IQR 33.0-35.9) and median HbA1c was 7.3% (IQR 6.7-7.8). The optimal model using physical activity achieved an R2 of 0.548 and a mean absolute error (MAE) of 30.30. The addition of electronic health record (ehr) information increased R2 to 0.636 and reduced MAE to 26.89 and the time of the day feature further increased R2 to 0.768 and reduced MAE to 20.55. Combining all the features together resulted in an optimal XGBoost model with an R2 of 0.836 and MAE of 17.47. This model accurately estimated normal glucose levels in 2584/2715 (95.2%) readings and hyperglycaemic events in 852/1031 (82.6%) readings, but fewer hypoglycaemic events (48/172 (27.9%)). The optimal XGBoost model prioritized age, gender, BMI and HbA1c followed by glucose levels and physical activity. Interestingly, the blood glucose level prediction by our model was influenced by use of SGLT2i.Conclusion: XGBoost, a machine learning AI algorithm achieves high predictive performance for normal and hyperglycaemic excursions, but has limited predictive value for hypoglycaemia in patients on multiple therapies who fast during Ramadan. [ABSTRACT FROM AUTHOR]

Given the surging growth of artificial-intelligence-inspired computational methods in materials science, experimental laboratories around the globe have become open to adopting data-driven approaches for materials discovery. The field witnesses emerging machine-learning models trained over databases, of which data are collected from high-throughput experimentation or first-principles calculation. Here, we address the impediment of constructing a highly accurate predictor for perovskite bandgap when the inorganic network undergoes the deformation. The predictor is trained on a dataset of first-principles calculations of pure and mixed-cation hybrid perovskites. We investigate the impact of the inclusion/exclusion of structural deformation features by training the model carefully. A high level of accuracy could be achieved with a scrupulous investigation of the input features. Our analysis emphasizes how important the feature selection is for the construction of the predictive model as we challenge the robustness of our machine learning predictor in a lab validation setup. [ABSTRACT FROM AUTHOR]