Welcome to KNUTS-DB

A data-driven knowledge database for Nuts containing information on chemical nutrients, molecular targets, pathways, disease associations and clinically defined allergen properties. The database contains data sets associated with extremely rich and diverse metadata on almond, cashew, pecan, hazelnut, walnut, pistachio and peanut.

The webpage is implemented as Shiny web-app written in the R programming language. It is hosted using a so-called shiny-server running on a Linux server within the Charité IT system. Shiny incorporates CSS definitions as well as javascript. The front end is implemented using the package “shinydashboard”. The Tables on the webpage were created with the package “DT”, an R-implementation of the well-known DataTables. The data is stored in a relational MySQL database. The database connection is handled through R using MySQL queries. For processing, depiction and data integration different R packages were used, most notably “tidyverse”, “ggplot2”, “UniprotR” and “webchem”. The web-app was tested on the most recent versions of Apple Safari, Google Chrome and Mozilla Firefox.

Data on individual allergens for respective nuts were identified from different databases and literature sources, and data on additional allergenic proteins are also reported in this section. One can search the database using respective nuts, allergen names, scientific names, physicochemical properties, genebank nucleotide, genebank protein ids, and UniProt ids. The filtered data can then be downloaded. Selecting single entries reveals the amino acid sequence, information about allergen-specific epitopes and cross-references.

The information about the nut composition was collected from different platforms: FoodData Central, fooDB, phenole explorer, Duke’s database and an in-house resource, superTCM. The data was then filtered for duplications by the InChI key and incomplete entries were removed. Additional data was obtained using an in-house text-mining pipeline using the PubMed query search and manual curation by domain experts. Furthermore, datasets were obtained and integrated from published literature sources. The chemical ingredients present in respective nuts can be searched using the name, IUPAC, SMILES, PubChem IDs, and protein interaction data. It is possible to click on an entry to show the corresponding data set, e.g. protein and pathway interactions.

The similarity search can be performt using either a name or a SMILES as query. If a name is used, the corresponding SMILES string is obtained from PubChem. The SMILES is than converted into a MACCS fingerprint and a Tanomoto similarity score for all compounds in the data base is calculated. The results are shown in a seperate table. It is possible to click on entries for further information about molecular functions, biological processes, cellular components and pathways of the protein.

The protein interaction data were derived from KEGG, BindingDB and ChEMBL by matching the compound InChI keys to database-specific IDs and using web scraping techniques. Since ChEMBL and BindingDB contains also data about binding experiments without successful binding, the data was then filtered for well-recorded binding entries, to receive a reliable dataset. Entries with inconclusive activity comments or comments indicating no binding were filtered out as well. Using information from UniProt, the data was restricted to reviewed proteins only. Protein interaction data on nuts can be retrieved using the protein-interaction tab of the database including information on UniPot id, enzyme names, protein name, organism, taxonomy and by interaction type and value. It is possible to click on entries for further information about molecular functions, biological processes, cellular components and pathways of the protein.

The platform allows the user to perform some basic exploratory data analysis. Therefore the user can choose data either from the whole data set of the platform or from the individual user defined data set. It is also possible to select data by interaction value type and to apply a threshold. In general, data can be selected by nut, compound, protein, pathway and interaction value type. The results of the analysis are returned as a plot and as an interactive data table. The pathway and gene ontology (GO-Term) enrichment analysis enables the user to identify over-represented pathways and GO-Terms in user-defined gene lists. Due to implementation related restrictions, the pathway and GO-term enrichment can only be performed for human proteins. The clustered heatmap plot is a valuable tool to inspect the compound protein relationship and identify interaction clusters, e.g. proteins that are affected by multiple compounds and vice versa. Hierarchical clustering with spearman distance is used and the coloring corresponds to the interaction value. The network plot is another way to inspect the compound protein relationship. Blue nodes represent compounds and orange nodes represent proteins. Compound and protein nodes are connected if there is an interaction. The size of the node corresponds to the number of distinct interactions and the thickness of the link to the interaction value (as thicker as lower the IC50, AC50, ki, ...). It is possible to zoom in and out and the nodes are movable to get a clear view on names and connections.

The part of the analysis that functioned as a starting point to this insight can now also be performed using our new platform. One possible approach is to filter for human pathways with interactions with peanut ingredients and a “Immune System” entry in the “Pathway Interaction'' page. Using “Add Data to user data set” a custom data set will be created which also includes the corresponding proteins and their interactions with nut ingredients. This data set is then selected in the “Network Plot'' page, restricted to IC50 values below 10,000nM to generate a network plot showing potential inhibitory influences (Figure 5). This plot reveals a cluster of fatty acids with a strong interaction with the Peroxisome proliferator-activated receptor alpha (PPAR-alpha), indicated by the thickness of the lines. It is now possible to inspect the corresponding data in the “Data”-tab of the “Network Plot” page, redistricting it to PPAR-alpha and to download the IC50 values if needed. For further information about the involved proteins, links to UniProt are provided. An overview about the pathways of PPAR-alpha can be found in the “Full data set” page using the corresponding columns to filter for PPAR-alpha, peanut and Homo sapiens. Here we can also find the Toll-Like Receptor Pathway 1 including further information and links.

Emanuel Kemmler, Margitta Worm, Robert Preissner and Priyanka Banerjee. KNUTS-DB - a data-driven knowledge database for NUTS. Toxicol Mech Methods 2025 Apr 29:1-8. DOI: 10.1080/15376516.2025.2496752.