TIN-X version 3: update with expanded dataset and modernized architecture for enhanced illumination of understudied targets

New in version 3.0

The most useful new feature in the improved UI is the creation of a switch allowing users to toggle between a plot (see Fig. 1) and a newly created table viewing mode (Figs. 2A, 2B). We added the new table view mode for both the Browse Diseases (Fig. 3A) and Browse Targets (Fig. 3B) TIN-X operational modes. This new feature allows users to view, sort, filter-search, and otherwise interact with all the information present in the log(Novelty) vs. log (Importance) scatterplots in a tabular format. In previous versions of TIN-X, this important target or disease data was not simultaneously accessible to the user in the scatterplot view as it now exists in the new table view mode. TIN-X Plot-view users can sort the table results ascending or descending by any of the fields and they can also perform a filter-search for results that immediately eliminates non-matching records while the user types the desired filter string. Users can click on any row in the new table view and bring up the same detailed popup that appears when a user clicks on a datapoint in the established plot view. Additionally, we have reformatted the multi-publication view to have consistent styling and to display the date of publication for each article. A browser compatibility issue was addressed regarding the slider located in the upper-right of the plot view (Fig. 2A) which is used to adjust the number of results to display. Also, when a user selects and deselects filter values while browsing results, the corresponding points in the scatterplot or table immediately disappear or appear based on the filter values being applied (Fig. 2A). Several other web browser compatibility issues and formatting inconsistencies were identified in the previous version of TIN-X and were subsequently corrected to improve the user experience. The UI has been tested with modern desktop versions of Google Chrome, Mozilla Firefox, Safari, IE/Edge, and Opera. Finally, the “About” section of TIN-X has been updated to describe version 3.0 of TIN-X with details such as when data sources were last updated and a list of contributors.

This new version of TIN-X features improvements to the TIN-X database. The TIN-X database now runs on Amazon RDS instead of mySQL v8 (Cannon et al., 2017). Additionally, changes were made to the TIN-X metadata table to improve performance. We also have a significantly expanded database due to Jensen Lab DISEASES resource version 2.0 indexing full-text articles from PubMed when available (Grissa et al., 2022) instead of only titles and abstracts. There are approximately nine-fold more target/disease associations in the new DISEASES 2.0 resource compared to previous versions (Grissa et al., 2022). This represents a meaningful increase in text-mined associations that improves the chances of biomedical researchers illuminating understudied targets.

The TIN-X API is accessible at api.newdrugtargets.org and is supported by Swagger documentation. The API has been adapted to support interaction with the new TIN-X Database in the cloud (Amazon Relational Database Service). Additionally, the entire API was upgraded from Python2 to Python3 in order to improve the reliability, stability, and interoperability of this resource and to extend the useful duration of the entire TIN-X application. The TIN-X API has been optimized to accommodate the expanded dataset. Like other components of TIN-X, several bugs that had been identified in the API are now fixed. A TIN-X REST API has also been integrated with the Common Fund Data Ecosystem (CFDE) Gene Pages Partnership Project.

In addition to existing as a standalone web application at https://newdrugtargets.org, TIN-X is now a featured resource within Pharos (https://pharos.nih.gov/), an NIH-hosted comprehensive drug target database and platform for exploring the druggable genome (Kelleher et al., 2023). As depicted in Fig. 4, text-mined target-disease associations from TIN-X are shown in an interactive scatterplot next to a circular treemap. The circular treemap groups the associations based on the hierarchy defined by the Disease Ontology (Schriml et al., 2022). Selecting a circle (group of diseases) or a point (individual disease) in the right panel highlights corresponding points on the scatterplot. The scatterplot is a log(importance) vs. log(novelty) plot, just like those produced by the standalone TIN-X web application. This successful integration of TIN-X with Pharos will help introduce Pharos users to the capabilities of TIN-X while providing an inline tool to visually explore disease associations.

Using TIN-X version 3.0

Users can Browse Diseases or Browse Targets by exploring the nested drop-down menus on the left or by typing the name of the target or disease of interest into the search/filter field in the top left. In the main plot area, each point represents a target associated with the disease selected in this example, bacterial infectious disease (Fig. 2A). These points are colored based on Target Development Level (TDL), and the shape of each point indicates the IDG Family, as indicated on the legend at the bottom (Fig. 2A). The next screenshot (Fig. 2A) shows a user searching for a specific Target among the many that exist. The autocomplete in this example suggests various Toll-like receptor targets, which if selected, highlight the corresponding point on the plot. In the upper-right, filters can be applied to add or remove points from the plot based on TDL category and/or IDG Family. The slider on the far right of this panel (Fig. 2A) appears in the upper right portion of the scatterplot view. Manipulating the slider either increases or decreases the number of associated diseases (or targets) that are shown. We order TIN-X target or disease results by their non-dominated solution (NDS) ranking optimizing both novelty and importance. By definition, NDS rank-one is assigned to solutions for which no solutions exist that are superior in all variables, NDS rank-two are solutions that satisfy that condition if rank-one solutions are removed, etc. By default, only the 300 most “interesting” results are displayed, based on NDS ranking. However, the user is free to adjust the number of results plotted using the slider depicted in Fig. 2A. In addition to the plot view, the new TIN-X introduces a powerful table view mode. The table displays all the same data available via the scatterplot but has advanced filtering capabilities, the ability to sort table results ascending or descending by any of the table’s fields, and a search filter that dynamically updates the displayed table results to exclude any rows that do not match the user-provided search string. Importantly, users can access the full details by clicking on any row (Fig. 3A), as was previously available by clicking on a point in the plot view. The expanded view contains details about the target and disease, an external link to Pharos, DrugCentral, a DTO ID, and the target’s family classification and TDL. This view also contains a list of the PubMed publications that form the basis for the predicted association between the target and disease of interest (Fig. 3A). Clicking the title of a publication in this list reveals the publication’s abstract, along with an external link to open the research article in PubMed.

Users can toggle back and forth between plot and table viewing modes as they explore the Browse Targets and Browse Diseases tools. Users can also generate a shareable link by using the Share option in the upper-right (Fig. 2A), and they can export a comma separated value (CSV) file of all the Targets or Diseases that are currently selected. In order to control the number of targets or diseases that are returned as results, the user can manipulate a slider control located in the upper right corner of the scatterplot. Clicking the “+” or “−” buttons increases or decreases the number of results. By default, all results are shown if there are less than 300 total, but if greater than 300 results exist, only the 300 most interesting results (those with the highest ranking NDS scores) are shown.

Use case: illumination of an understudied target for Parkinson’s disease

Parkinson’s disease (PD) is a chronic, degenerative disorder of the central nervous system primarily affecting the motor system (Kalia & Lang, 2015). Both environmental and genetic factors (Warner & Schapira, 2003; Kalia & Lang, 2015; Lill, 2016) likely contribute to the development of PD in an individual. PD is a complex and intensely studied disease that resists more traditional, manual techniques for conducting relevant literature searches due to the vast and growing body of peer-reviewed PD publications. TIN-X is well suited to discover and explore PD-associated targets with an emphasis on illuminating understudied targets.

The Browse Diseases mode of TIN-X is used to explore understudied targets associated with PD. The list of diseases on the left navigation of TIN-X or the search option can be used to locate Parkinson’s Disease. The “Parkinson’s Disease” category contains three subcategories in the TIN-X Browse Diseases mode: Late-onset PD, Early-onset PD, and Juvenile-onset PD. Selecting the parent term, “Parkinson’s disease” in the diseases navigation shows results for all three forms of PD. Figure 5A shows the resulting table view of targets associated with PD, by default ordered by the NDS rank of Importance and Novelty scores. Examination of this table (Fig. 5A) reveals several T_dark targets among the top targets ordered by increasing nds_rank.

Among the top-ranked T_dark targets (19th overall in the table) is the T_dark target ATP10B, which TIN-X predicts to have high PD-relevance. The highlighted row in Fig. 5A is the understudied target ATP10B, which encodes a late endosomal lipid flippase (Timcenko et al., 2019). Figure 5B depicts the same targets in Plot mode on a log(Importance) vs. log(Novelty) scatterplot with the T_dark target ATP10B highlighted via mouseover. Clicking on this datapoint in the scatterplot (or alternatively clicking the corresponding row in the Table-view) reveals details about ATP10B and PD including external links to Pharos and DrugCentral as shown in Fig. 5C. The six articles responsible for TIN-X’s predicted association between ATP10B and PD are listed here (Fig. 5C).

A noteworthy article in this list proposes that mutated forms of ATP10B increases the risk of PD by compromising lysosomal glucosylceramide export (Martin et al., 2020). This is the first study to demonstrate that ATP10B encodes a late endosomal lipid flippase which is responsible for translocating glucosylceramide and phosphatidylcholine from the luminal to the cytosolic membrane leaflet. The authors also conducted genetic screening among cohorts of patients with early-onset PD and dementia with Lewy bodies and identified seven total patients as heterozygous mutation-carriers of the ATP10B P4-Type ATPase gene (Martin et al., 2020). They show in isolated neurons that the loss of ATP10B is responsible for lysosomal dysfunction and associated cell death which is important because lysosomal dysfunction is known to be involved in PD pathology. The results of this study suggest that “recessive loss-of-function mutations in ATP10B increase the risk for PD via disturbed lysosomal export of glucosylceramide and phosphatidylcholine” (Martin et al., 2020) Specifically, the authors make a case that loss-of-function mutations in ATP10B lead to reduced lysosomal concentrations of glucosylceramide which in turn causes lysosomal dysfunction. In this example, TIN-X leads the user to this significant recent research, which proposes specific mechanisms by which an understudied target, ATP10B, might be implicated in PD.

Other articles and associated correspondence are listed among the TIN-X article results for ATP10B and PD, including: “ATP10B variants in Parkinson’s disease: a large cohort study in Chinese mainland population” (Zhao et al., 2021), and also the correspondence “Reply: ATP10B variants in Parkinson’s disease: a large cohort study in Chinese mainland population” (Smolders & Van Broeckhoven, 2021). The authors of this article (Zhao et al., 2021) and associated correspondence (Smolders & Van Broeckhoven, 2021) were motivated by (Martin et al., 2020) but did not find convincing statistical evidence among the cohorts they examined linking mutations in ATP10B to PD. However, due to limitations in these population studies, further investigation is needed to better assess the relationship between genetic mutations in ATP10B and PD. As shown in Fig. 5C, the addition of article publication dates and improved formatting in Version 3.0 of TIN-X helps the user explore the collection of articles responsible for the predicted association between ATP10B and PD. In this case, the chronology of associated articles is important because the later studies (Smolders & Van Broeckhoven, 2021; Zhao et al., 2021) cite the earlier, groundbreaking work linking ATP10B and PD (Martin et al., 2020). Users exploring the list of articles can rapidly understand the basis of TIN-X’s prediction by clicking on an article title, revealing the abstract for that article within the TIN-X UI and providing an external link to the full-text article.

In this use-case, TIN-X is shown to aid in illuminating the understudied target ATP10B by assigning high Importance and Novelty scores to this T_dark target while providing a convenient interface for exploring the research articles responsible for this predicted association.