Case study: NCI’s Molecular Targets Platform

Case Studies Oct 31, 2023

This blog post is part of a series that will explore applications and expansions of the Open Targets informatics ecosystem, particularly the Open Targets Platform and Open Targets Genetics through conversations with our users.

The National Cancer Institute (NCI)’s Childhood Cancer Data Initiative (CCDI) is supporting the development of an instance of the Open Targets Platform specific to childhood cancers for systematic drug target identification and prioritisation. The CCDI Molecular Targets Platform (MTP) is a collaborative effort between the Children’s Hospital of Philadelphia and the Frederick National Laboratory with input from the NCI and the U.S. Food and Drug Administration (FDA).

The goal of MTP is to integrate the FDA’s Relevant Molecular Target List and enrich the platform with pre-clinical paediatric cancer datasets containing somatic variation, gene expression, and gene fusion data to aid in identifying new paediatric cancer-specific drug targets. Data integration under common ontological terms will aid in categorising diseases, genes, and agents.

The MTP is is a clinical community-driven example of the many potential applications of the Open Targets resources open-source code to create customised instances that can integrate patient-specific, experimental or pre-publication results — with the overarching aim to facilitate therapeutic hypotheses building and target discovery for the entire community.

What was your motivation for creating this resource?

[Deanne Taylor, lead Principal Investigator of MTP] In the US, formal clinical trials are required to establish the safety and utility of those treatments in pediatric cancers. Despite the evidence that pediatric cancers could benefit from the same drugs used to treat adult cancers, drug trials are much rarer in children than adults. To address this disparity, the USA’s Research to Accelerate Cures and Equity (RACE) for Children Act as of 2020 requires any company designing a clinical trial for an antineoplastic therapy to include a pediatric trial component if the gene target is on the FDA’s Pediatric Molecular Target Lists (PMTL). While the PMTL broadly suggests genes based on expert reviews and literature, the actual types and subtypes of pediatric cancers that could be targeted are not described. By collecting and harmonizing molecular data across all available pediatric cancer research studies, we wanted to provide the evidence of particular gene targets within as many types of pediatric cancers as possible. We envisioned this resource supporting the stated use by companies, but also use by pediatric cancer researchers, pediatric oncologists and the public.

Why did you choose Open Targets as your backbone infrastructure?

[Deanne Taylor] The purpose of the project was to deliver harmonized genomic information to support therapeutic discovery in pediatric cancer. The Open Targets Platform had several advantages as a backbone infrastructure for this purpose. Open Targets delivers easy-to-navigate aggregated information and evidence linking therapies, genes, and diseases. Open Targets is generally a familiar platform in industry, which reduces the barrier for use. As pediatric cancers are believed to have a germline component, Open Targets has the advantage that it also integrates genetic data, including genetic studies of pediatric cancer, through Open Targets Genetics.

OpenPedCan and Molecular Targets Platform Data Sources

The Open Pediatric Cancer (OpenPedCan) Project at the Children's Hospital of Philadelphia (CHOP) is an open analysis effort performing downstream analysis on harmonised childhood cancer data from multiple sources. A few examples of datasets that have been added to MTP from OpenPedCan are:

Like the Open Targets Platform, the MTP currently integrates the Experimental Factor Ontology (EFO), with a plan to add NCI Thesaurus and other resources to enable the platform to better support target discovery.

Additionally, the team has developed the R API framework for paediatric cancer customised visualisations found within the MTP - like their non-batch corrected Differential Expression Heatmap interactive widget for both Target and Disease pages — as shown in the examples below:

KRAS profile page in the MTP, showing the pediatric cancer widgets (in green) side-to-side to the Open Targets ones

Differential Expression heatmap widget in a Gene Page (ex. KRAS), where one gene in all pediatric diseases vs all GTex tissue is displayed

The team are actively developing the MTP, releasing v2.1 earlier this year. “We are excited to continue adding data, visuals, and functionality to the MTP so that researchers, clinicians, and the entire pediatric oncology community can more easily mine the data to identify new cancer types to target with existing therapies or molecular targets as drugs and trials are being developed.”

What are the main challenges you faced in the MTP development stage?

[Yizhen Chen, Development Lead] We created the Molecular Targets Platform by building on and extending the Open Targets capabilities to prioritize molecular targets relevant in pediatrics, adolescents, and young adults (AYAs) based on preclinical data.

The Open Targets Platform is a sophisticated system. From a developer perspective, the Open Targets team did a great job on data harmonization, data sharing, and transparency of the ecosystem.

However, the complexity of the system requires a lot of domain knowledge and software engineering experience, and the major challenge for us sits on the data side: it is challenging to ensure the data, configuration settings, or scripts we are using are updated at the same time as the Open Targets Platform. Validating the output of the ETL process and monitoring compatibility and synchronicity between the two platforms is not always straightforward.

In particular, it can take us months to re-implement the MTP features if the architecture design of the Open Targets Platform changes. To make this easier, we have set up a GitHub repo called starter-kit, with a general overview of how all MTP project repositories connect to each other, with guidelines on the proper way to use them. The Open Targets team is also actively working on platform-output-support and is now sharing their technical roadmaps in advance to each release, which will simplify the infrastructure setup.

Photo of Yizhen Chen, with the caption: Yizhen is the Development Lead for the Molecular Targets Platform. Quote: I appreciate the opportunity to make contributions to the community around childhood cancer care and research data. I love to see that our product can improve our understanding of cancer biology.

What are your plans for MTP going forwards?

[Subhashini Jagu, NCI Federal Lead and Mark Cunningham, Technical Project Manager] We are exploring several options to expand and evolve the platform, including supporting data from multiple providers, elaborating the searchable information related to FDA's Pediatric Molecular Target Lists, and ways to create a more streamlined and sustainable operational model for MTP. We also plan to support reusability of our data, and are therefore very happy to hear that the Open Targets team is planning to integrate our pediatric dataset into the Platform.

Photo of Subhashini Jagu with the caption: Subhashini is the NCI Federal Lead Chief, Scientific Policy and Program Branch A Chief, Office of Data Sharing, CBIIT, NCI. Quote: I love to work on MTP because this resource allows investigators to validate FDA's Pediatric Relevant Molecular Targets List, thereby complementing the Research to Accelerate Cures and Equity for Children Act (RACE) and contributing to the important mandate for pediatric investigation by any industry partners developing drugs that target those variants.

Photo of Mark Cunningham, with the caption: Mark is the Technical Project Manager for the Molecular Targets Platform. Quote: The reasons I love being part of the MTP project is the privilege of working with an amazing team — NCI, Frederick National Laboratory's engineers, and CHOP — and contributing, in some small way, to our ability to improve treatment outcomes for childhood and AYA cancers through enhanced target discovery and data sharing.

How do you think your resource can help build therapeutic hypotheses for childhood cancer & address drug discovery questions?

[Deanne Taylor/The team] The most common forms of childhood cancer can be further classified by specific molecular subtypes. Clinicians decide on an optimal treatment regimen based on the patient’s cancer subtype. There may be additional pediatric cancer subtypes that may benefit from particular therapies, but those subtypes have not yet been identified because of rarity or lack of evidence. Identifying new subtypes and developing therapies for subtypes requires a concerted effort from investigators who can follow leads and patterns in molecular evidence from one or more genomics studies. To support discovery of rarer subtypes and their therapeutic approaches, this project provides researchers with harmonised (co-processed) data that would allow for rapid integration of various datasets.

The team concluded by saying: “We believe that through enhanced data sharing, we can improve our understanding of cancer biology so that new preventative measures and treatments may be uncovered. Our goal is to ensure that researchers learn from every child with cancer in order to extend the survivorship and quality of life for children with pediatric cancers”