Velsera

Dennis A. Dean

In the context of rapidly evolving fields such as genomics and personalized medicine, how do you adapt your bioinformatics strategies to meet the changing demands of the industry?

Advancing drug discovery has become my single focus in that it is a way to align all of our efforts toward what I see as the most impactful goal. Genomics and Personalized medicine is a marathon that starts with early discovery that identifies physiological mechanisms associated with promising drug targets. Years later, with billions of dollars of investment, new treatments can be available to alleviate patient suffering. Our focus in the life sciences has to be reducing the cost and time to get new life-altering treatments to patients. Consequently, we aim to align each bioinformatics project so that it adds to the tooling required to accelerate drug discovery.

My initial consideration is to be clear on bioinformatics' impact on the broader life sciences. We aim to identify potentially reusable modules, the availability of open-source workflows, and the potential to develop generalizable tools for other projects. Common Workflow Language (CWL) and Nextflow have been critical in our ability to reuse, scale, and empower the community to advance research and the drug discovery pipeline. Workflows implemented in our systems are reproducible, which can assist in the transition to regulatory submissions.

What are some challenges in integrating bioinformatics into the research and development new treatments, drugs, or therapies in the life sciences?

Planning for Quality Control Procedures in Large Scale Analysis. Large-scale multi-omics analysis across institutions can introduce analysis challenges in practice. Some of these challenges include differences in sequencing technology, workflow processing, and clinical/functional annotation variations. Even when many aspects of data generation are controlled, batch effects can be an issue. We have found integrating quality control procedures essential in all multi-omics analyses during the development and for production workflow. Identifying issues as they arise can allow teams to minimize potential downstream analysis in interpretation while saving time and cost.

Multi-Omics Analysis Standards. Preparing data for large-scale processes can take just as much, if not more effort, than processing large data sets and can require substantial expertise. This is especially true when tools are not developed in-house or when harmonizing data across institutions. Multi-omics standards can potentially decrease the effort required to harmonize and combine data while ensuring information needed to interpret datasets is maintained throughout the analysis process.

Bridging Clinical Efficacy Gap. Estimates show that as low as 10% of drugs transition from phase 1 studies to new treatments. Reasons for a new drug's failure can include toxicity, efficacy, and bioavailability. Moreover, bioinformatics is just one component of drug discovery. Frameworks that integrate failure databases (negative results), bioinformatics, and chem-informatics with large preclinical and clinical data sets could allow for guided identification or exclusion of targets earlier in the drug discovery pipeline.

What is your vision for the future of translational science and bioinformatics?

Public private partnerships to reduce cost and time: My vision of the future of translational science and bioinformatics is the continued commitment to reduce costs and time to new treatments through competitive collaborations. My experience in working with public-private partnerships is very promising. Some challenges that can be addressed when collaborating include developing and operationalizing standards to conduct multi-model analysis, identifying minimum technical data elements, assembling standardized databases, and developing standardized methods, including workflows. The goal of these partnerships must be to create a common platform for which to explore novel approaches. 

If we can reduce the effort required to conduct standardized analysis, we can spend more time analyzing results and conducting novel analyses that may provide new insights. Reducing the time to conduct standard analyses is a major feature of our platforms. However, it is not enough. We need collaborations from early discovery to impacting patients to leverage what has come before, especially regarding bioinformatics. My experience is that workflows are re-developed at multiple transitions to account for implementation preferences or adapt to data structure differences. The more we can work as a field to execute standards, the more we can leverage past efforts. This will become even more important as multi-omics and multi-modal analysis fuel the precision medicine pipeline.

Racial and ethnic best practices: I would love to see racial and ethnic considerations integrated into all precision medicine projects because it is a best practice. Data-driven projects must have information regarding data diversity. This will become increasingly important as we amass large datasets as a community. It is much easier to collect diverse information at the start of a project. We must be intentional about knowing the exact status of the data used in our bioinformatics and infer, if not test, what the implications may be.

Revolutionary computational approaches, including the Insilco experiment: The ATOM consortium is another of my favorite projects. ATOM is a public-private partnership aimed at democratizing drug discovery. The ATOM Modeling Pipeline contains specialized data databases, cheminformatics tooling, and deep learning functionality to advance drug discovery. I can see a future where widely available drug discovery tools will augment bioinformatics. I would expect that bioinformatics will both inform and be informed by a wider availability of drug discovery tools.

National data and computational ecosystems facilitate feedback to precision medicine approaches: A national data and computational Ecosystem would allow for data discovery across institutions, allowing for harmonization and partitioning in support of evaluation, analysis, and the development of precision medicine tools. Computational analysis could leverage local resources to access computational analysis. The system would be aware of data access restrictions, and security measures would be built into the Ecosystem. The Ecosystem would also facilitate sharing previous medicine best practices, allowing for best practices to be quickly shared across institutions.

Do you have any advice, suggestions, or warnings you would like to give to professionals in a similar role working in the pharmaceutical industry?

Bioinformatics project team building: It is not uncommon for bioinformatics teams to include 10 to 20 people, and each person has a specific role in ensuring the project's success. To build and interpret workflow output, deep scientific, biochemistry, and clinical knowledge may be required. Care should be taken to ensure select team members are transdisciplinary trained and able to consult with multiple disciplines.

Focus on relationships: Knowing the history of bioinformatics implementation is essential, especially when workflows are shared with collaborators and when they become a community standard. Building a network of individuals and teams that understand the details of bioinformatic workflows is essential for systematically resolving issues as they arise. Developing a network of workflow experts also ensures that the community is robust to inevitable changes in personnel.