The Chemical Shift Index or CSI 3.0 (http://csi3.wishartlab.com) is a web server designed to accurately identify the location of secondary and super-secondary structures in protein chains using only nuclear magnetic resonance (NMR) backbone chemical shifts and their corresponding protein sequence data. Unlike earlier versions of CSI, which only identified three types of secondary structure (helix, β-strand and coil), CSI 3.0 now identifies total of 11 types of secondary and super-secondary structures, including helices, β-strands, coil regions, five common β-turns (type I, II, I′, II′ and VIII), β hairpins as well as interior and edge β-strands. CSI 3.0 accepts experimental NMR chemical shift data in multiple formats (NMR Star 2.1, NMR Star 3.1 and SHIFTY) and generates colorful CSI plots (bar graphs) and secondary/super-secondary structure assignments. The output can be readily used as constraints for structure determination and refinement or the images may be used for presentations and publications. CSI 3.0 uses a pipeline of several well-tested, previously published programs to identify the secondary and super-secondary structures in protein chains. Comparisons with secondary and super-secondary structure assignments made via standard coordinate analysis programs such as DSSP, STRIDE and VADAR on high-resolution protein structures solved by X-ray and NMR show >90% agreement between those made with CSI 3.0.

Protein chemical shifts have long been used by NMR spectroscopists to assist with secondary structure assignment and to provide useful distance and torsion angle constraint data for structure determination. One of the most widely used methods for secondary structure identification is called the Chemical Shift Index (CSI). The CSI method uses a simple digital chemical shift filter to locate secondary structures along the protein chain using backbone (13)C and (1)H chemical shifts. While the CSI method is simple to use and easy to implement, it is only about 75-80% accurate. Here we describe a significantly improved version of the CSI (2.0) that uses machine-learning techniques to combine all six backbone chemical shifts ((13)Cα, (13)Cβ, (13)C, (15)N, (1)HN, (1)Hα) with sequence-derived features to perform far more accurate secondary structure identification. Our tests indicate that CSI 2.0 achieved an average identification accuracy (Q3) of 90.56% for a training set of 181 proteins in a repeated tenfold cross-validation and 89.35% for a test set of 59 proteins. This represents a significant improvement over other state-of-the-art chemical shift-based methods. In particular, the level of performance of CSI 2.0 is equal to that of standard methods, such as DSSP and STRIDE, used to identify secondary structures via 3D coordinate data. This suggests that CSI 2.0 could be used both in providing accurate NMR constraint data in the early stages of protein structure determination as well as in defining secondary structure locations in the final protein model(s). A CSI 2.0 web server (http://csi.wishartlab.com) is available for submitting the input queries for secondary structure identification.

The recent Coronavirus Disease 2019 (COVID-19) pandemic has put a tremendous burden on global health systems. Medical practitioners are under great pressure for reliable screening of suspected cases employing adjunct diagnostic tools to standard point-of-care testing methodology. Chest X-rays (CXRs) are appearing as a prospective diagnostic tool with easy-to-acquire, low-cost and less cross-contamination risk features. Artificial intelligence (AI)-attributed CXR evaluation has shown great potential for distinguishing COVID-19-induced pneumonia from other associated clinical instances. However, one of the associated challenges with diagnostic imaging-based modeling is incorrect feature attribution, which leads the model to learn misguiding disease patterns, causing wrong predictions. Here, we demonstrate an effective deep learning-based methodology to mitigate the problem, thereby allowing the classification algorithm to learn from relevant features. The proposed deep-learning framework consists of an ensemble of convolutional neural network (CNN) models focusing on both global and local pathological features from CXR lung images, while the latter is extracted using a multi-instance learning scheme and a local attention mechanism. An inspection of a series of backbone CNN models using global and local features, and an ensemble of both features, trained from high-quality CXR images of 1311 patients, further augmented for achieving the symmetry in class distribution, to localize lung pathological features followed by the classification of COVID-19 and other related pneumonia, shows that a DenseNet161 architecture outperforms all other models, as evaluated on an independent test set of 159 patients with confirmed cases. Specifically, an ensemble of DenseNet161 models with global and local attention-based features achieve an average balanced accuracy of 91.2%, average precision of 92.4%, and F1-score of 91.9% in a multi-label classification framework comprising COVID-19, pneumonia, and control classes. The DenseNet161 ensembles were also found to be statistically significant from all other models in a comprehensive statistical analysis. The current study demonstrated that the proposed deep learning-based algorithm can accurately identify the COVID-19-related pneumonia in CXR images, along with differentiating non-COVID-19-associated pneumonia with high specificity, by effectively alleviating the incorrect feature attribution problem, and exploiting an enhanced feature descriptor.

Applications of machine learning algorithms (MLAs) to modeling the adsorption efficiencies of different heavy metals have been limited by the adsorbate–adsorbent pair and the selection of specific MLAs. In the current study, adsorption efficiencies of fourteen heavy metal–adsorbent (HM-AD) pairs were modeled with a variety of ML models such as support vector regression with polynomial and radial basis function kernels, random forest (RF), stochastic gradient boosting, and bayesian additive regression tree (BART). The wet experiment-based actual measurements were supplemented with synthetic data samples. The first batch of dry experiments was performed to model the removal efficiency of an HM with a specific AD. The ML modeling was then implemented on the whole dataset to develop a generalized model. A ten-fold cross-validation method was used for the model selection, while the comparative performance of the MLAs was evaluated with statistical metrics comprising Spearman’s rank correlation coefficient, coefficient of determination (R2), mean absolute error, and root-mean-squared-error. The regression tree methods, BART, and RF demonstrated the most robust and optimum performance with 0.96 ⫹ R2 ⫹ 0.99. The current study provides a generalized methodology to implement ML in modeling the efficiency of not only a specific adsorption process but also a group of comparable processes involving multiple HM-AD pairs.

Accessible surface area (ASA) is the surface area of an atom, amino acid or biomolecule that is exposed to solvent. The calculation of a molecule's ASA requires three-dimensional coordinate data and the use of a "rolling ball" algorithm to both define and calculate the ASA. For polymers such as proteins, the ASA for individual amino acids is closely related to the hydrophobicity of the amino acid as well as its local secondary and tertiary structure. For proteins, ASA is a structural descriptor that can often be as informative as secondary structure. Consequently there has been considerable effort over the past two decades to try to predict ASA from protein sequence data and to use ASA information (derived from chemical modification studies) as a structure constraint. Recently it has become evident that protein chemical shifts are also sensitive to ASA. Given the potential utility of ASA estimates as structural constraints for NMR we decided to explore this relationship further. Using machine learning techniques (specifically a boosted tree regression model) we developed an algorithm called "ShiftASA" that combines chemical-shift and sequence derived features to accurately estimate per-residue fractional ASA values of water-soluble proteins. This method showed a correlation coefficient between predicted and experimental values of 0.79 when evaluated on a set of 65 independent test proteins, which was an 8.2 % improvement over the next best performing (sequence-only) method. On a separate test set of 92 proteins, ShiftASA reported a mean correlation coefficient of 0.82, which was 12.3 % better than the next best performing method. ShiftASA is available as a web server ( http://shiftasa.wishartlab.com ) for submitting input queries for fractional ASA calculation.