A Useful Guide to Lectin Binding: Machine-Learning Directed Annotation of 57 Unique Lectin Specificities.
Bojar D, Meche L, Meng G, Eng W, Smith DF, Cummings RD, Mahal LK. · 2022
Read A Useful Guide to Lectin Binding: Machine-Learning Directed Annotation of 57 Unique Lectin Specificities. by Bojar D, Meche L, Meng G, Eng W, Smith DF, Cummings RD, Mahal LK. (2022) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc9679999.txt.
Introduction
Bojar and co-authors create a guide to lectin binding, with machine learning notes on 57 lectin specificities. Lectins are proteins that grip sugar chains called glycans. Most glycan knowledge rests on what these probes bind. Their exact tastes were poorly defined for years.
The team used a large glycan microarray data set as their base. The arrays tested trade lectins against hundreds of glycans. They blended 68 hand-set glycan features with machine learning rules. Expert notes then joined the model output for each lectin. The guide sorts the 57 lectins by their lead binding taste. Mannose binders and complex N-glycan binders lead the groups. Fucose, sialic acid, and GalNAc binders follow in turn. The authors say the guide shows how tangled lectin tastes can be. Labs can now pick probes with clearer eyes.
Catalog record
| CREATOR | Bojar D, Meche L, Meng G, Eng W, Smith DF, Cummings RD, Mahal LK. |
|---|---|
| YEAR | 2022 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 104 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Useful Guide to Lectin Binding: Machine-Learning Directed Annotation of 57 Unique Lectin Specificities By Daniel Bojar, Lawrence Meche, Guanmin Meng, William Eng, David F Smith, Richard D Cummings, Lara K Mahal Published in: ACS Chemical Biology DOI: 10.1021/acschembio.1c00689 ABSTRACT Abstract Glycans are critical to every facet of biology and medicine, from viral infections to embryogenesis. Tools to study glycans are rapidly evolving; however, the majority of our knowledge is deeply dependent on binding by glycan binding proteins (e.g., lectins). The specificities of lectins, which are often naturally isolated proteins, have not been well-defined, making it difficult to leverage their full potential for glycan analysis. Herein, we use a combination of machine learning algorithms and expert annotation to define lectin specificity for this important probe set. Our analysis uses comprehensive glycan microarray analysis of commercially available lectins we obtained using version 5.0 of the Consortium for Functional Glycomics glycan microarray (CFGv5). This data set was made public in 2011. We report the creation of this data set and its use in large-scale evaluation of lectin–glycan binding behaviors. Our motif analysis was performed by integrating 68 manually defined glycan features with systematic probing of computational rules for significant binding motifs using mono- and disaccharides and linkages. Combining machine learning with manual annotation, we create a detailed interpretation of glycan-binding specificity for 57 unique lectins, categorized by their major binding motifs: mannose, complex-type N-glycan, O-glycan, fucose, sialic acid and sulfate, GlcNAc and chitin, Gal and LacNAc, and GalNAc. Our work provides fresh insights into the complex binding features of commercially available lectins in current use, providing a critical guide to these important reagents. INTRODUCTION Carbohydrates (i.e., glycans) are involved in every facet of life from the cell walls of bacteria to the signals that start inflammatory cascades in humans.1 While our understanding of other biomolecules, such as DNA and RNA, has expanded exponentially due to the advent of new analytical technologies, glycans have remained understudied due to the lack of convenient analytical tools. Nature solved the problem of glycan identification by designing noncatalytic glycan binding proteins called lectins, which recognize well-defined epitopes within a glycan. Lectins, which can be found in all organisms, are often isolated for commercial use from plants and have long been used as tools for analysis of the mammalian glycome.2 The connection between the biological functions of plant lectins and their binding of mammalian epitopes is not well understood. However, these probes have commonly been used to probe glycosylation. Our earliest understanding of blood group antigens comes from agglutination studies using lectins to determine blood type.3 More modern methods that leverage lectins as analytical tools include lectin histology and enzyme-linked lectin assays (ELLA).3 Lectin microarrays, in which a panel of lectins (10s to >100) and other carbohydrate-binding probes (e.g., antibodies) are printed on a solid support, are now routinely used in glycomics.4−6 This high-throughput method has identified glycans involved in a variety of systems from melanoma metastasis to viral host response.7−9 Although lectins are useful tools, they have suffered from a lack of detailed definition of their binding requirements, which hampers their analytical utility. Traditional methods of defining lectin specificity have involved inhibition assays with monosaccharides, binding assays with a limited set of potential ligands, or in a few cases, crystal structures with disaccharide or monosaccharide binders. These are often the data commercial suppliers of lectins provide about binding specificities, typically pointing to mono- or disaccharide motifs. In more recent years, glycan microarrays, in which the binding of probes to hundreds of glycans are interrogated simultaneously, have been used to identify more detailed glycan-binding specificities.10−12 In 2011, the Mahal Laboratory, in collaboration with the Consortium for Functional Glycomics (CFG), collected glycan microarray data for 116 commercially available lectin preparations using version 5.0 of the Consortium for Functional Glycomics glycan microarray (CFGv5). This data set was made public via the CFG database in 2011, and subsets of this data have been used to perform cross-platform comparisons of glycan microarrays,13,14 model lectin-glycan interactions,15 and create new bioinformatic methods for glycan microarray analysis.16−19 Herein, we analyze this data set, using a combination of machine learning and expert manual annotation, to provide a useful guide to the glycan-binding specificities of these commercial probes.13 Connecting binding events with glycan substructures or motifs has attracted considerable interest for the purpose of identifying glycan-receptor interactions. Related work has focused on analyzing lectins and their cognate binding motifs from a protein sequence or structural perspective.20,21 Under the assumption that similar sequences and/or structures bind similar ligands, lectins can be grouped into classes with potentially shared binding properties. This can aid lectin categorization, annotation, and utility for researchers working with these lectins. This concept has recently been taken further with LectinOracle,22 a deep learning algorithm that utilizes protein and glycan sequences to predict lectin specificity. To annotate the specificities of glycan binding proteins more directly from glycan microarray data, multiple algorithms including frequent subtree mining18,23 and motif based approaches have been developed.16,17 Yet the high diversity and nonlinearity of glycans has stymied the large-scale evaluation of subtle, interpretable binding motifs in glycan array data to date. To overcome this, we have leveraged the recent introduction of machine learning into glycobiology.24 By mapping inputs (glycan sequences) to outputs (lectin-glycan binding), machine learning algorithms can ascertain the most important features (i.e., sequence motifs) that predict lectin–glycan binding. Importantly, this is performed on a scale that is vastly larger than manual annotation and also enables the analysis of highly complex feature combinations to obtain insights into more subtle influences of the co-occurrence of glycan motifs. We further engaged in feature engineering by combining hand-crafted features that are domain-relevant (e.g., the presence of Lewis A or terminal sialosides) with systematic probing for all observed sequence motifs of various lengths. This procedure improves the interpretability of the resulting machine learning models, which have traditionally been labeled as uninterpretable “black boxes”. We then used these features, our machine learning models, and iterative manual annotation to establish logical, interpretable rules for each lectin that best explained lectin–glycan binding behavior to further facilitate “white box” machine learning and extract lectin binding specificities. Upon removal of duplicate lectins (e.g., preparations of the same lectin from different sources) and lectins with poor binding, our analysis provides detailed annotation of 57 unique commercially available lectins. While lectin binding specificities are routinely analyzed on the basis of glycan array data, the most common mode of analysis is to explain what is being bound, for instance, via shared or enriched motifs. Here, we expand on this concept by also considering glycans that are not bound by a lectin, both in our machine learning analysis and in the subsequent expert annotation. This procedure allowed us to identify additional binding determinants that, for instance, inhibit binding to a preferred motif when present and overall improve the precision and usefulness of our lectin binding annotation. The results of our work give a clear view into the binding profiles of these carbohydrate-binding reagents and demonstrate that they are selective in their epitope binding. Overall, our work provides a useful guide to anyone analyzing glycans using this reagent class and sets the stage for more advanced interpretation of studies using lectins. RESULTS AND DISCUSSION Study