CanImmunother is a up-to-date manually curated database for cancer immunotherapy. Currently, CanImmunother provides 3267 experimentally supported associations between 218 cancer sub-types across 34 body parts and 484 immunotherapies with 642 biomarkers and 108 targets through manual curation on 4515 publications. CanImmunother provides user-friendly web applications and web APIs to browse, search, analyse, and access data freely for precision cancer immunotherapy.
Figure. 1 The framework of CanImmunother.
The CanImmunother database has been extensively tested by beta-testers, at different common browsers (such as Firefox 68.0, Google Chrome 61, Safari, and Microsoft Edge 88.0.705.50) at personal computer with Windows system and smart phone with different operation systems (such as Android 10 and IOS 12.5.1).
2.1 Browse:
The “Browse all” web-page presents all of the associations in a table to allow users to filter for the interesting associations through the “search” box (Figure 2 A). Moreover, we implemented ‘Browse’ sub-web-pages to allow users to browse interesting entries through different body parts, cancer sub-types, biomarker types, and immunotherapy types (Figure 2 B). In addition, at the “Home” web-page, the symbols in the word-cloud diagrams are hoverable to show its numbers of entries, and clickable to browse entry details (Figure 2 C).
Figure 2. The web interface of browse application.
2.2 Search:
The search application allows users to search by cancer sub-type, immunotherapy, biomarker, or target with setting filtration parameters, such as body part, immunotherapy type, biomarker type, and research type (Figure 3 A). The quickly search box in the resulted table is to allow users to filter out interesting associations efficiently (Figure 3 B). And the "details" button links to further webpages for more information of the association in publication. In addition, CanImmunother provides external links to the related reference resources, such as NCBI PubMed, EFO, DrugBank, NCIT OBO Edition, UniprotKB, and ECO (Figure 3 C). Moreover, CanImmunother provides the expression profile(s) of immunotherapy associated target(s) across 15 body parts in the GTEx and TCGA resources (Figure 3 D).
Figure 3. The web interface of search application.
2.3 Excel table application:
The excel table application was implemented to help users to find interesting associations efficiently from large numbers of associations. For example, a simple search can obtained 297 associations for non-small cell lung carcinoma (NSCLC) (Figure 4). By using the excel table application for combinational filtration, four associations, which describe atezolizumab versus docetaxel with their biomarkers and targets for NSCLC, was efficiently found from the 297 associations (Figure 4). Moreover, we can further access these four associations in CanImmunother to compare the clinical efficacy and adverse event of atezolizumab immunotherapy with docetaxel chemotherapy in NSCLC (Figure 4).
Figure 4. The web interface of excel table application.
2.4 Association prioritization:
To promptly prioritize the associations, an association prioritization application was developed based on the non-redundant association data in CanImmunother. The association prioritization application allows users to retrieve a cancer sub-type, an immunotherapy, a biomarker, and a target alone or together for prioritizing cancer immunotherapy association data with their predictive biomarkers and targets (Figure 5 A). For instance, we searched the keywords like “Neutrophil-to-lymphocyte ratios” and “Renal cell carcinoma” together. And we promptly found that “Neutrophil-to-lymphocyte ratios” is a common biomarker for renal cell carcinoma patients with nivolumab, pembrolizumab, ipilimumab, atezolizumab, and avelumab treatment to target PD-1 and CTLA-4 (Figure 5 A). And the most associated immunotherapy of “Neutrophil-to-lymphocyte ratios” is nivolumab for renal cell carcinoma (Figure 5 A). In addition, each prioritizing association can optionally visualized in a network (Figure 5 B).
Figure 5. The web interface of association prioritization application.
2.5 Network visualization:
To explore the relationships among different cancer sub-types, immunotherapies, biomarkers, and targets, a network visualization application was implemented in CanImmunother. The application allows users to input a set of cancer sub-types, immunotherapies, biomarkers, and targets, and to construct interactive networks to display their relationships. As the interactive networks may consist of many nodes and even more interactions (Figure 6), CanImmunother offers users filtering function to hide those interactions that are less interesting. The filtering function allows users to filter out nodes and interactions by the node name of cancer sub-types, immunotherapies, biomarkers, and targets. When selecting or unselecting some of the nodes, the interactive network will be changed accordingly, and a sub-network of the whole interactive network will be displayed. Moreover, the association data displaying in the network diagram can be download for further analysis in house. In addition, all nodes in the network allow adjustment except for node legends.
Figure 6. The web interface of network visualization application.
2.6 Other applications:
CanImmunother also offers the RESTful APIs for the programmatical access of the association data. The resulted data by the APIs are available in universal JSON and text formats. Documentation for the APIs is available on the website. All association data in CanImmunother can be freely downloaded. In addition, CanImmunother encourages users to submit their new association data for future data integration. Once checked by our professional curators and approved by the submission review committee, the submitted records will be included in an updated release.
In CanImmunother, the cancer immunotherapy associations are manually collected and curated from different publications. Referring to the association score models used in ncRNAVar (PMID:33275967), ncRPheno (PMID:32122231), and Open Targets (PMID:27899665), CanImmunother refined an association score model to prioritize the associations between cancer sub-types and immunotherapies with their biomarkers and targets based on two evidential metrics. These evidential metrics include the research types and the number of the supporting publications. The computation of the association score consists of three steps as below:
Step 1: CanImmunother assigned different weights for the different research types of publications according to their quality. In Equation (1), the association was curated from case report, W is originally assigned as 0.4; the association was curated from basic research publications, W is originally assigned as 0.6; the association was curated from clinical research publications, W is originally assigned as 1.0.
Step 2: As a larger number of publications can enhance the evidential value (E) for a same association. A harmonic sum function (PMID:33275967, PMID:32122231, PMID:27899665) was used to compute E of an association with multiple supporting publications but also to dampen the effect of a large number of publications. E is calculated as Equation (2). In Equation (2), ‘n’ represents the total number of supporting publications, and W1, W2, W3, …, Wn are sorted in a descending order.
Step 3: The E above are normalized to limit the range of final association scores (Score) from 0 to 1.0. In Equation (3), ‘e’ represents the natural constant e.
The following information can help you to interpret cancer immunotherapy associations in CanImmunother.
Cancer sub-type
To make the cancer symbols more consistent with other databases, CanImmunother provides both the identifier and link for the cancer sub-type in the Experimentally Factor Ontology (EFO) resource.
Immunotherapy
To make the immunotherapy symbols more consistent with other databases, CanImmunother provides both the identifier and link for the immunotherapy in the DrugBank database and NCI Thesaurus OBO Edition(NCIT) resource.
Biomarker
To make the biomarker more consistent with other databases, CanImmunother provides both the identifier and link for the biomarker in the UniprotKB database and NCI Thesaurus OBO Edition(NCIT) resource.
Target
To make the target more consistent with other databases, CanImmunother provides both the identifier and link for the target in the UniprotKB database.
If you have any inquiries, please do not hesitate to contact us via email: zhangwl25@mail3.sysu.edu.cn