Actinobacteria TFDB: An integrated view of transcription factors in Actinobacteria

Abstract

Actinobacteria represent a prolific source of bioactive natural products. However, the complex transcriptional regulatory networks in these bacteria, particularly the interplay between transcription factors (TFs) and their regulatory ligands (TF-RLs), remain poorly characterized and lack dedicated resources. In this context, we introduce the Actinobacteria Transcription Factor Database (Actinobacteria TFDB), a comprehensive repository that systematically integrates TF-centric data across 25 representative species. The current version encompasses 629 TFs, classified into 69 families, documents 11,776 TF-target relationships and 28 TF posttranslational modification sites. Uniquely, it features a dedicated collection of 54 experimentally validated TF-RL interactions. Beyond providing standardized annotations, sequence and structural features, and regulatory networks, Actinobacteria TFDB incorporates a specialized TF-RL module that enables interactive exploration and visualization of allosteric regulatory mechanisms. By consolidating multi-dimensional TF data from diverse sources, this resource empowers systems-level analyses and facilitates the rational design of regulatory strategies to activate silent biosynthetic gene clusters and optimize metabolite production. The database is publicly available at http://mingleadgene.com:9315/#/home.

Graphical abstract

1. Introduction

Actinobacteria, a major phylum of Gram-positive bacteria, are distinguished by their sophisticated morphological differentiation and remarkable capacity for producing bioactive secondary metabolites, thereby representing a cornerstone of pharmaceutical discovery and industrial biotechnology [[1], [2], [3]]. Among them, the genus Streptomyces epitomizes the concept of a “natural drug factory” with individual genomes frequently harboring >50 biosynthetic gene clusters (BGCs) dedicated to secondary metabolism [[4], [5], [6]]. However, the inability to activate >60% of these BGCs under conventional cultivation conditions remains a significant obstacle. This phenotypic silence is primarily dictated by multi-layered transcriptional regulatory circuits [7].

The remarkably high abundance of transcription factors (TFs) in actinobacterial genomes represents a fundamental genetic determinant underlying their exceptional capacity for secondary metabolism [[8], [9], [10], [11]]. In the model actinobacteria Streptomyces coelicolor, which possesses a large genome of approximately 8000 genes, an estimated 965 to 1232 genes are predicted to encode TFs, accounting for 12% to 15% of its total gene complement [12]. This proportion is more than double that observed in typical bacteria, such as Escherichia coli, where TFs constitute approximately 5% to 7% of the genome, and rivals the regulatory complexity of certain simple eukaryotes, including Drosophila [[13], [14], [15]]. These transcription factors form elaborate and hierarchical regulatory networks, wherein a single regulator can direct the expression of hundreds of target genes, thereby synchronizing morphological differentiation with the production of antibiotics [8,[16], [17], [18], [19], [20], [21]]. From a biotechnological perspective, rational engineering of key TFs, via strategies including overexpression or targeted point mutations, has proven highly effective in amplifying antibiotic titers, often by several- to several dozen-fold [17,[22], [23], [24], [25]]. Notable examples include the rewiring of pathway-specific regulators in industrial strains, which has yielded over 50-fold enhancements in the biosynthesis of clinically relevant compounds such as avermectin and daptomycin [[26], [27], [28]]. Collectively, these findings underscore the central importance of TFs in governing the biology of actinobacteria and in driving their industrial exploitation for natural product discovery and optimization.

In addition, TFs are modulated by diverse regulatory ligands, including small-molecule effectors (such as nucleotide second messengers, metabolic intermediates, ions and antibiotics) and protein regulatory ligands [29]. Their binding induces allosteric changes or alters protein interactions, fine-tuning transcriptional responses [30,31]. Small molecules primarily sense environmental and metabolic states, whereas protein regulatory ligands enhance complex assembly and stability [[24], [30], [31]]. This integrated signaling enables the transcriptional reprogramming necessary for adaptation, governing processes from metabolism to antibiotic resistance [[32], [33], [34], [35], [36]]. The centrality of regulatory ligands highlights the multi-layered nature of regulation in Actinobacteria, demanding holistic study through multi-dimensional data that encompass TF sequences, DNA-binding motifs, and regulatory ligand interactions.

While high-throughput methods such as ChIP-seq and ATAC-seq have enabled genome-wide mapping of TF networks in Actinobacteria [37,38], revealing regulatory circuits for antibiotic production and virulence in species such as Saccharopolyspora erythraea and Mycobacterium tuberculosis [39,40], the field remains hindered by data fragmentation. TF functional data are dispersed across studies with inconsistent annotations and available databases are inadequate for integrated analysis. This fragmentation ultimately obstructs a system-level understanding of transcriptional regulation and its application in metabolic engineering.

Here, we developed the Actinobacterial Transcription Factor Database (Actinobacteria TFDB). This database consolidates foundational TF information from sources such as Ensembl, KEGG and UniProt with multi-dimensional features, including experimentally validated TF regulatory ligand interactions. It provides a unified platform for analyzing TF-target networks and regulatory modifications, thereby supporting diverse research aims from elucidating transcriptional mechanisms to guiding natural product discovery and synthetic biology applications. Additionally, the database serves as a reference for industrial strain design via multi-omics data integration. This article outlines the construction pipeline, data sources, and core functionalities of Actinobacteria TFDB, which offers both a roadmap for its ongoing development and a robust, standardized resource for the research community.

2. Materials and methods

2.1. Data collection and curation

The data within Actinobacteria TFDB were systematically curated from multiple sources to ensure comprehensiveness and accuracy. Primary data on TFs, including species and family classification, were collected from scientific literature databases (PubMed [41], Web of Science) and functional annotation databases (KEGG database [42], PubChem [43], UniProt [44], NCBI GenBank [41]). To achieve structured storage of standardized data, we developed a closed-loop pipeline for Actinobacteria transcription factor data: first, manual curation built a seed dataset of basic annotations; then, using KEGG gene IDs, a Python crawler retrieved related information from NCBI and UniProtKB, effectively resolving identifier mappings across databases. After cleaning and standardization including generating hyperlinks to external resources, the data was exported as an Excel file conforming to the MySQL table schema. Finally, a Java program mapped Excel headers to database fields for batch warehousing, with support for exception logging and resumable uploads.

The quality of the integrated data was ensured through rigorous cross-verification and manual scrutiny of information from the various sources. To ensure data quality, we implemented a dual-validation approach combining automated screening with manual review, where customized SQL statements were first used to detect anomalies across multi-dimensional data, followed by manual verification, confirmation, and correction of flagged entries. This process ensured data uniqueness and completeness by validating the uniqueness of transcription factor records based on species, TF name, and gene ID; verifying target gene uniqueness by species and gene ID; confirming TF family classification integrity through record counts; assessing species coverage breadth; and ensuring uniqueness of TF-regulatory ligand associations via a quadruple of TF name, ligand name, species, and external ID. The final curated dataset provides multi-dimensional information for each TF, which is organized and presented through several core modules on the website. These primarily include Basic Information, Transcript details, and TF-RL interactions. For instance, using a known KEGG identifier, users can access a dedicated information page with links to external resources like KEGG, enabling them to obtain deep functional insights such as family classification, protein sequence, and involvement in biochemical pathways. Similarly, searches using an Ensembl ID or a UniProt Swiss-Prot ID provide direct links to the respective databases, granting access to key sequence data, structural details, and domain characteristics of the TFs. By integrating these diverse categories of information, Actinobacteria TFDB offers users a specific and comprehensive overview of their TF of interest (Fig. 1).

Fig. 1.

Data processing workflow of Actinobacteria TFDB. The data processing workflow in Actinobacteria TFDB includes four main parts: (top)Multi-source data collection and standardized verification; (left) Data processing and integration, including Excel-based input, automated SQL generation with GO term annotation, and final database deployment via a CentOS 7.9 server; (bottom) Detailed information display, presenting basic information, transcript details, and TF-RL; (right) User function modules, supporting keyword-based and category-based searches.

2.2. Database contents

Actinobacteria TFDB is a comprehensive, specialized resource dedicated to the TFs of Actinobacteria. The current release systematically integrates data across 25 species, encompassing 629 TFs, 69 TF families, 11,776 TF-target relationships, 28 TF posttranslational modification sites and a unique collection of 54 experimentally validated TF-RL interactions (Table 1). The integrated dataset is accessible through a user-friendly web interface with the following core pages, each designed to facilitate specific research queries and data exploration.

Table 1.

Coverage of TFs in the Actinobacteria TFDB database.

Species	TF numbers
Mycobacterium smegmatis	292
Mycobacterium tuberculosis H37Rv	182
Mycolicibacterium smegmatis MC2 155	1
Saccharopolyspora erythraea NRRL 2338	11
Saccharopolyspora erythraea NRRL 2341	1
Saccharopolyspora pogona NRRL30141	7
Saccharopolyspora spinosa	1
Streptomyces avermitilis ATCC31267	10
Streptomyces avermitilis KA320	3
Streptomyces chartreusis NRRL 3882	1
Streptomyces coelicolor A3(2)	82
Streptomyces cyanogenus S136	1
Streptomyces griseus subsp. griseus NBRC 13,350	6
Streptomyces griseus [Yersinia enterocolitica subsp. palearctica Y11]	1
Streptomyces lincolnensis	5
Streptomyces lydicus NRRL 2433	1
Streptomyces pactum Act12	2
Streptomyces pristinaespiralis	1
Streptomyces pristinaespiralis NRRL2958	4
Streptomyces rapamycinicus	1
Streptomyces roseosporus SW0702	1
Streptomyces sp ATCC 55,186	1
Streptomyces venezuelae	2
Streptomyces venezuelae ATCC 10,712	11
Streptomycessp. CS40	1

The Home page serves as the central gateway to the database. It provides a concise overview of the biological significance of Actinobacterial TFs and the database's core content. A prominent and convenient search function is placed directly on this page, allowing users to immediately initiate queries by entering a TF name, species, or database identifier (Ensembl ID, UniProt ID). It ensures that both new and returning users can quickly access the core functionality of the database.

The dedicated Search page offers two versatile and complementary modes for data retrieval (Fig. 2a). The "Search by Keyword" mode supports flexible querying by letting users enter any transcription factor-related identifier in a unified input box. For more targeted investigations, the "Search by Category" mode enables users to filter data through multiple biologically relevant criteria. (Fig. 2b). The search results are presented in a sortable table with key annotation information (Fig. 2c). Each entry in the results table includes a "View" button, which directs the user to a dedicated detail page for comprehensive information (Fig. 2d).

Fig. 2.

Search interface of Actinobacteria TFDB. (a) Main search interface. (b) Interface for category-based search with filtering options. (c) Sortable summary table for displaying matched search results. (d) Dedicated page for presenting complete annotation information of individual transcription factors.

In the Species page, we classify TFs by species and construct a tree diagram, which illustrates the associations between different Actinobacteria species and various TF families (Fig. 3a). Users can access information on TFs from 25 Actinobacteria species in this section, and clicking on relevant nodes allows viewing of corresponding details. Following this is the percentage of family members section, which uses a pie chart to present the proportion of different TFs, with families such as TetR and LysR each having their corresponding proportion values (Fig. 3b). This design enables users to intuitively understand the composition of TFs in specific species and facilitates in-depth exploration of species-specific characteristics related to transcriptional regulation in Actinobacteria.

Fig. 3.

Visualization of transcription factor classification and family distribution. (a) Association graph on the Species page: First-level nodes represent Actinobacterial species; selection of a species node reveals second-level nodes of associated TF families. (b) Proportional distribution of TF families, visualized as an arc chart on the Species page. (c) Hierarchical tree diagram on the TF page: First-level nodes denote TF families, which expand into second-level nodes of individual TFs upon selection. (d) Multi-level circular chart on the TF page: The outermost ring shows the relative abundance of all TF families; interaction with the corresponding region navigates detailed views of family and member composition.

The TF page provides a global overview of all transcription factors within the database. It features an interactive association graph that visually maps the connections between Actinobacteria species and their diverse TFs. TFs are systematically classified into a tree diagram by family; selecting any family node expands the structure to reveal its constituent TFs, and choosing a specific TF prompts a summary information box to appear below (Fig. 3c). Complementing this, a hierarchical pie chart intuitively displays the distribution of TF families across the dataset (Fig. 3d). The initial view shows the proportional representation of each family. Selecting a specific sector drill down to a second level, which details the selected family's composition, while a third level is accessible by selecting an individual TF. The page concludes with the "Basic Domains Group" section, which dynamically presents all TFs in a tabular format, including Ensembl ID, Gene ID, Family, and Species. A "View" button, available both in the summary pop-up and the table, directs users to the dedicated detail page for any TF. Collectively, this page facilitates systematic exploration of the distribution patterns and characteristics of TFs from different families, enabling users to filter and retrieve relevant information based on their transcription factor of interest.

A defining feature and core advantage of the Actinobacteria TFDB is its integration of experimentally validated TF-RL interactions. This type of curated functional data remains relatively scarce in existing databases, yet it is crucial for deciphering the sophisticated regulatory networks that govern Actinobacterial physiology and secondary metabolism [45]. As summarized in Table 2, the database currently documents over 30 specific TF-RL pairs across 8 species, providing a unique resource for investigating signal perception and allosteric regulation. The compiled TF-RL interactions involve a diverse array of key regulatory molecules, which can be categorized into several functional classes, including cyclic nucleotide second messengers, key metabolic intermediates, and antibiotic-derived signaling molecules [[46], [47], [48]].

Table 2.

Coverage of known TF-RLs in the Actinobacteria TFDB database.

Species	TF	TF-RL
Mycobacterium smegmatis	EtbR	ethambutol [49]
Mycobacterium smegmatis	Lsr2	cyclic di-3′,5′-guanylate [50]
Mycobacterium smegmatis	XbpR	D-xylose [51]
Saccharopolyspora erythraea NRRL 2338	DasR	c-di-AMP [52]
Saccharopolyspora erythraea NRRL 2338	BldD	c-di-GMP [53]
Saccharopolyspora erythraea NRRL 2338`	GlnR	acetylated GlnA1 [54]
Mycobacterium tuberculosis H37Rv	Rv0273c	ethambutol [49]
Mycobacterium tuberculosis H37Rv	Rv1473A	adenosine 5′-triphosphate [55]
Mycobacterium tuberculosis H37Rv	Rv2640c	cupric cation [56]
Streptomyces venezuelae ATCC 10,712	RmdA	c-di-GMP [57]
Streptomyces venezuelae ATCC 10,712	RmdB	c-di-GMP [57]
Streptomyces griseus subsp. griseus NBRC 13,350	ArpA	cAMP [58]
Streptomyces avermitilis	AccR	methylcrotonyl-CoA [59]
Streptomyces avermitilis	AccR	propionyl-CoA [59]
Streptomyces avermitilis	AccR	acetyl-CoA [59]
Streptomyces avermitilis	SAV4189	hygromycin B [60]
Streptomyces avermitilis	SAV4189	thiostrepton [60]
Streptomyces lincolnensis	SLCG_Lrp	arginine [47]
Streptomyces lincolnensis	SLCG_Lrp	phenylalanine [47]
Streptomyces avermitilis	AveT	avermectin B1 [48]
Streptomyces coelicolor A3(2)	AdpA	sulfane sulfur [61]
Streptomyces coelicolor A3(2)	BldD	cyclic di-3′,5′-guanylate [62]
Streptomyces coelicolor A3(2)	BldD	nitric oxide [62]
Streptomyces coelicolor A3(2)	CatR	hydrogen peroxide [63]
Streptomyces coelicolor A3(2)	CdaR	tryptophan [64]
Streptomyces coelicolor A3(2)	Crp	cAMP [65]
Streptomyces coelicolor A3(2)	DasR	N-Acetyl-d-glucosamine [66]
Streptomyces coelicolor A3(2)	GluR	glutamic acid [67]
Streptomyces coelicolor A3(2)	HypR	zinc ion [68]
Streptomyces coelicolor A3(2)	HypR	L-hydroxyproline [68]
Streptomyces coelicolor A3(2)	SCO3361	L-phenylalanine [10]
Streptomyces coelicolor A3(2)	SCO3361	L-cysteine [10]
Streptomyces coelicolor A3(2)	NmtR	nickel (2+) [69]
Streptomyces coelicolor A3(2)	NmtR	cobalt (2+) [69]
Streptomyces coelicolor A3(2)	NsrR	nitric oxide [70]
Streptomyces coelicolor A3(2)	OxyR	hydrogen peroxide [71]
Streptomyces coelicolor A3(2)	ScbA	Streptomyces gamma-butyrolactone 1 [72]
Streptomyces coelicolor A3(2)	ScbR	Streptomyces gamma-butyrolactone 1 [72]
Streptomyces coelicolor A3(2)	ScbR2	jadomycin B [73]
Streptomyces coelicolor A3(2)	ScbR2	actinomycin D [74]
Streptomyces coelicolor A3(2)	ScbR2	undecylprodigiosin [74]
Streptomyces coelicolor A3(2)	SlbR	Streptomyces gamma-butyrolactone 1 [40]
Streptomyces coelicolor A3(2)	SoxR	phenazine methosulfate [75]
Streptomyces coelicolor A3(2)	SoxR	p-Benzoquinone [75]
Streptomyces coelicolor A3(2)	SoxR	plumbagin [75]
Streptomyces coelicolor A3(2)	SoxR	diamide [75]
Streptomyces coelicolor A3(2)	StgR	γ-actinorhodin [76]
Streptomyces coelicolor A3(2)	TamR	citrate [77]
Streptomyces coelicolor A3(2)	TamR	trans-aconitate [76]
Streptomyces coelicolor A3(2)	TamR	cis-aconitate [76]
Streptomyces coelicolor A3(2)	TamR	1-hydroxypropane-1,2,3-tricarboxylic acid [76]
Streptomyces coelicolor A3(2)	XdhR	guanosine 3′,5′-bis(diphosphate) [78]
Streptomyces coelicolor A3(2)	XdhR	guanosine3′-diphosphate 5′-triphosphate [78]
Streptomyces coelicolor A3(2)	XdhR	guanosine 5′-triphosphate [79]

To maximize the utility of this unique dataset, we developed a dedicated TF-RL search module. This module is specifically designed to store, retrieve, and display the high-quality, experimentally validated interaction data highlighted in Table 2. It integrates auxiliary structural and functional annotations from authoritative external databases, including the Protein Data Bank (PDB) for 3D structural context and UniProt for sequence and domain information, thereby constructing a comprehensive data framework for each interaction.

To enable efficient and diverse querying of the curated TF-RL interactions, the module features a user-centric search system operating through two complementary modes: a flexible keyword search for direct queries using TF names, regulatory ligand names, or species, and a targeted category search that allows for filtering across predefined biological dimensions such as TF Name, TF Family, TF-RL Type, and Source Species (Fig. 4a, b). The category search interface is streamlined with "Submit" and "Reset" buttons for effortless refinement. Search results are presented in a paginated table, clearly displaying key columns including TF Name, TF Family, TF-RL Name, and Source Species, with a crucial "View" button in the "Details" column (Fig. 4c). Clicking this button navigates to a dedicated, comprehensive information page for the specific TF-RL pair. This page features a structured layout centered on a "Basic Information" section that integrates extensive data, typically encompassing TF details, regulatory ligand information, interaction context, and direct cross-references to external databases, allowing users to access a holistic functional profile without needing to navigate multiple external platforms (Fig. 4d).

Fig. 4.

TF-RL interaction search interface of Actinobacteria TFDB. (a) Search interface supporting both keyword and category-based queries. (b) Category search panel with filters for TF name, TF-RL name, and species. (c) Search results are listed in a paginated table; the “View” button leads to a dedicated detail page. (d) The detail page provides comprehensive information, including TF and regulatory ligand names, source species, and links to external databases such as UniProt and PDB.

Unlike general resources such as GTRD [80] and KnockTF 2.0 [81], which are eukaryote-centric, or specialized tools like RiboD [82] and RSwitch [83] that focus only on riboswitches, Actinobacteria TFDB provides comprehensive Actinobacteria-specific TF data. Compared to prokaryote-focused databases such as CollecTF [84], LogoMotif [85], and RegPrecise [86], our platform integrates TF sequences, TF–RL interactions, target genes, and posttranslational modification sites (Table 3). Notably, RegPrecise [86] lacks key Actinobacteria regulators like PhoP [87] and BldD [88], limiting its utility for systematic studies in this phylum. These comparisons highlight the unique and integrative value of Actinobacteria TFDB.

Table 3.

Comparison of transcriptional regulation-related databases.

Database	Species	Focus Area	Regulatory Ligand Query	Core Functions
RiboD	Prokaryotes	Riboswitches	None	Predict, annotate and visualize prokaryotic riboswitches
RSwitch	Prokaryotes	Riboswitches	None	Analyze riboswitch sequence/structure; evaluate drug targets for pathogenic bacteria
GTRD	Eukaryotes	TF and transcription factor binding sites (TFBS)	None	Integrate ChIP-seq/DNase-seq data; predict and analyze eukaryotic TFBS
KnockTF 2.0	Eukaryotes	TF, TF-RL and target genes	Available	Integrate knockdown/knockout data; annotate and analyze TF target genes
CollecTF	Prokaryotes	TF and TFBS	None	Integrate bacterial TFBS experimental data; conduct motif alignment and analysis
LogoMotif	Actinobacteria	TF, TFBS and target genes	None	Predict and visualize actinobacteria TFBS; support BGC regulatory locus analysis
RegPrecise	Prokaryotes	TF, TF-RL and TF-target genes	Available	Classify, visualize and analyze conservation of prokaryotic TF
Actinobacteria TFDB	Actinobacteria	TF, TF-RL and TF-target genes and posttranslational modification sites	Available	TF annotation and regulatory network visualization

3. Discussion

Actinobacteria TFDB was created to integrate the fragmented landscape of transcriptional regulation data in Actinobacteria. Through the systematic consolidation of TF data across 25 species, it directly tackles the challenge of data dispersion and the inadequacy of existing general resources. All TF-target gene relationships in this study were defined based on systematic literature mining and are supported by published experimental evidence. The database strategically encompasses key model species such as Streptomyces coelicolor and Mycobacterium tuberculosis, thereby providing support for antibiotic development and research into host-pathogen dynamics. A key feature of this database is its collection of experimentally validated TF-RL interactions, which offers molecular-level insight into signal integration in Actinobacteria. Furthermore, the systematic annotation of TF families and orthologs enables systems-level analyses previously constrained by data fragmentation. These resources support rational approaches to strain engineering, with potential applications in developing antimicrobials and other high-value compounds.

Actinobacteria are known for their production of diverse bioactive compounds, and their genomes contain a vast reservoir of silent biosynthetic gene clusters (BGCs). Activating these BGCs through rational manipulation of transcriptional regulators is a key objective in the field [89]. Actinobacteria TFDB could serve as a platform for this purpose by mapping the master transcriptional switches of BGCs and the regulatory ligands that regulate them. For instance, a researcher studying antibiotic production in Streptomyces can search for the TF BldD to obtain its regulatory ligands, target genes involved in secondary metabolism, and associated pathway information. This allows identification of potential targets for genetic engineering to enhance antibiotic yield. Another example: users can input a target gene such as SCO5085 (actII-ORF4, a pathway-specific activator in actinorhodin biosynthesis) to perform a reverse query and identify upstream TFs regulating its expression. Additionally, by integrating transcriptional network data with homology information, researchers can predict regulatory mechanisms for uncharacterized TFs in nonmodel strains. These applications demonstrate how the database supports hypothesis-driven experimental design in metabolic engineering and natural product discovery.

Actinobacteria TFDB remains a work in progress with several directions planned for its development. Incremental updates will be performed annually with a formal version and data package released. New KEGG genome data will be integrated, TF-target relationships supplemented through literature mining, and TF family classifications updated according to international standards. Each update undergoes format, logic, and manual verification. Future versions will broaden taxonomic representation, integrate ChIP-seq and other functional genomic datasets, and offer user-friendly tools for binding site and network analysis. Community feedback will be essential for expanding TF-RL interaction data, validating predicted functions, and maintaining annotation accuracy. Together, these efforts will ensure the database continues to serve as a unified, species-specific platform for studying Actinobacterial transcriptional regulation. By centralizing dispersed data and providing accessible interfaces, Actinobacteria TFDB aims to accelerate research in microbial genetics, natural product biosynthesis, and strain engineering.

Data availability statement

The web interface to the database is available at http://mingleadgene.com:9315/#/home.

CRediT authorship contribution statement

Yu Fu: Writing – original draft, Investigation, Funding acquisition, Formal analysis. Zhan-Hui Xu: Writing – original draft, Investigation, Formal analysis. Yi-Fan Liang: Writing – original draft, Investigation, Formal analysis. Shi-Qi Yang: Investigation, Formal analysis. Xue-Qin Xie: Investigation, Formal analysis. Bang-Ce Ye: Visualization, Validation, Supervision, Funding acquisition. Di You: Writing – original draft, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.