Machine Learning on Signal-to-Noise Ratios Improves Peptide Array Design in SAMDI Mass Spectrometry

Discovery of Bioactive Peptides Through Peptide Scanning

Therapeutic peptides targeted at various diseases are becoming increasingly relevant for the pharmaceutical industry. Several of these drugs were originally designed by mimicking a segment of a protein of interest. As such, protein mimicry represents a promising strategy both in immunology, for the identification of B- and T-cell epitopes, as well as for the modulation of protein activity, including the disruption of protein-protein interactions (PPIs) and the interference with biological or pathological cellular functions. Several methods have been developed to pinpoint the (binding) epitopes of a protein or the regions responsible for biological activity. One of such strategies is the scanning of the protein or selected domains with synthetic overlapping peptides. As the mechanism of action of a mimetic peptide can be similar to that of the whole protein, this method offers a powerful tool for the investigation of protein function, along with providing a solid basis for the development of therapeutic candidates. This review gives a general overview of different applications of the peptide scanning methodology, describing a comparison of the preparation and use of solid-phase libraries (peptide arrays) with isolated peptide libraries and highlighting their strengths and most common applications.

Machine-Learning Classification of Bacteria Using Two-Dimensional Tandem Mass Spectrometry

Biothreat detection has continued to gain attention. Samples suspected to fall into any of the CDC's biothreat categories require identification by processes that require specialized expertise and facilities. Recent developments in analytical instrumentation and machine learning algorithms offer rapid and accurate classification of Gram-positive and Gram-negative bacterial species. This is achieved by analyzing the negative ions generated from bacterial cell extracts with a modified linear quadrupole ion-trap mass spectrometer fitted with two-dimensional tandem mass spectrometry capabilities (2D MS/MS). The 2D MS/MS data domain of a bacterial cell extract is recorded within five s using a five-scan average after sample preparation by a simple extraction. Bacteria were classified at the species level by their lipid profiles using the random forest, k-nearest neighbor, and multilayer perceptron machine learning models. 2D MS/MS data can also be treated as image data for use with image recognition algorithms such as convolutional neural networks. The classification accuracy of all models tested was greater than 99%. Adding to previously published work on the 2D MS/MS analysis of bacterial growth and the profiling of sporulating bacteria, this study demonstrates the utility and information-rich nature of 2D MS/MS in the identification of bacterial pathogens at the species level when coupled with machine learning.

APD optimal bias voltage compensation method based on machine learning

The signal-to-noise ratio (SNR) of avalanche photodiode (APD) in optical detection system is greatly influenced by background radiation and operating temperature, so an APD optimal bias voltage compensation method based on machine learning is designed to accurately judge the current laser emission state and APD working state so that dichotomy compensation can be carried out to make APD work in optimal state. By means of cross-verification, the accuracy of judging laser emission state and APD working state is as high as 100% and 99.3% separately, then the number of input variables in the model is reduced appropriately by experiment and the prediction speed of the algorithm is further improved. Finally, road detection application is taken as the experimental background and comparison between the proposed method and the most widely used signal amplitude feedback compensation method is carried out. The results of this study suggest that the proposed APD optimal bias voltage compensation method based on machine learning offers a new and promising approach.

Profiling Protein Tyrosine Phosphatase Specificity with Self-Assembled Monolayers for Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry and Peptide Arrays

The opposing activities of phosphatases and kinases determine the phosphorylation status of proteins, yet kinases have received disproportionate attention in studies of cellular processes, with the roles of phosphatases remaining less understood. This Research Article describes the use of phosphotyrosine-containing peptide arrays together with matrix-assisted laser desorption/ionization (MALDI) mass spectrometry to directly profile phosphatase substrate selectivities. Twenty-two protein tyrosine phosphatases were characterized with the arrays to give a profile of their specificities. An analysis of the data revealed that certain residues in the substrates had a conserved effect on activity for all enzymes tested, including the general rule that inclusion of a basic lysine or arginine residue on either side of the phosphotyrosine decreased activity. This insight also provides a new perspective on the role of a R1152Q mutant in the insulin receptor, which is known to exhibit a lower phosphorylation level and which this work suggests may be due to an increased activity toward phosphatase enzymes. The use of self-assembled monolayers for matrix-assisted laser desorption/ionization mass spectrometry (SAMDI-MS) to provide a rapid and quantitative assay of phosphatase enzymes will be important to gaining a more complete understanding of the biochemistry and biology of this important enzyme class.

Intelligence Algorithms for Protein Classification by Mass Spectrometry

Mass spectrometry (MS) is an important technique in protein research. Effective classification methods by MS data could contribute to early and less-invasive diagnosis and also facilitate developments in the bioinformatics field. As MS data is featured by high dimension, appropriate methods which can effectively deal with the large amount of MS data have been widely studied. In this paper, the applications of methods based on intelligence algorithms have been investigated. Firstly, classification and biomarker analysis methods using typical machine learning approaches have been discussed. Then those are followed by the Ensemble strategy algorithms. Clearly, simple and basic machine learning algorithms hardly addressed the various needs of protein MS classification. Preprocessing algorithms have been also studied, as these methods are useful for feature selection or feature extraction to improve classification performance. Protein MS data growing with data volume becomes complicated and large; improvements in classification methods in terms of classifier selection and combinations of different algorithms and preprocessing algorithms are more emphasized in further work.