Phen2Disease: a phenotype-driven model for disease and gene prioritization

Impact summary

The introduction of Human Phenotype Ontology (HPO)-based approaches has brought about significant advancements in the field of genomic diagnostics for rare diseases. However, these approaches often underutilize the vast amount of available information concerning disease and patient phenotypes. In our study, we present a groundbreaking method called Phen2Disease, which effectively prioritizes diseases and genes by employing bidirectional maximum matching semantic similarity between patient and disease phenotype sets. Our comprehensive experiments encompassed six real data cohorts and two simulated data cohorts, totalling 2051 cases. The results demonstrated the superior performance of Phen2Disease compared to three state-of-the-art methods, particularly in cohorts with a lower average number of HPO terms. Furthermore, we observed that patients with higher information content scores possessed more specific information, resulting in more accurate predictions. Phen2Disease not only achieves exceptional accuracy but also provides remarkable interpretability through ranked diseases and patient HPO terms. This novel method offers a promising approach to leveraging phenotype data for genomic diagnostics of rare diseases, potentially leading to significant clinical impact. Researchers can access Phen2Disease freely on GitHub at https://github.com/ZhuLab-Fudan/Phen2Disease.

Research and engagement activities leading to impact

Our research collected real-world datasets and introduced an innovative methodology called Phen2Disease, designed to enhance the prioritization of diseases and genes by leveraging bidirectional maximum matching semantic similarity between patient and disease phenotype sets. Through our investigations, we discovered a strong correlation between higher information content scores in patients and the presence of more precise information, thereby facilitating more accurate predictions. Additionally, Phen2Disease offers exceptional interpretability, presenting ranked diseases alongside patient HPO terms. This novel approach to utilizing phenotype data in genomic diagnostics for rare diseases holds promising potential for substantial clinical impact.

Research outputs associated with the impact

1. A real-world dataset:
The research project produced a valuable real-world dataset that serves as a foundational resource for further investigations and analyses in the field. The dataset encompasses comprehensive and reliable data collected from diverse sources, providing a rich and representative sample of real-world cases. It offers researchers the opportunity to explore, analyze, and draw meaningful conclusions based on empirical evidence. This dataset contributes to the advancement of knowledge and facilitates evidence-based decision-making in various domains.

2. A new approach called Phen2Disease:
As part of the research project, a novel approach called Phen2Disease was developed, which represents a breakthrough in the field. Phen2Disease is an innovative methodology that leverages advanced computational techniques and algorithms to link phenotypic information with disease associations. By integrating and analyzing vast amounts of complex data from multiple sources, Phen2Disease identifies meaningful patterns, correlations, and potential causal relationships between phenotypes and diseases. This approach has the potential to advance the field by providing new insights into disease mechanisms, facilitating early diagnosis, and guiding personalized treatment strategies.

3. The published paper in the top journal: Briefings in Bioinformatics:
The research findings were disseminated through a published paper in the prestigious journal Briefings in Bioinformatics, showcasing the significance and impact of the research. Publication in such a top-tier journal underscores the rigorous scientific methodology employed in the study and signifies its contributions to the scientific community. The paper presents a comprehensive and detailed account of the research project, including the research objectives, methodologies, experimental results, and insightful conclusions. Its inclusion in Briefings in Bioinformatics ensures wide visibility, attracting the attention of researchers, professionals, and experts in the field. This publication plays a crucial role in advancing knowledge, stimulating further research, and inspiring future studies in the domain.

Researcher involvement

Our research endeavors have involved a dedicated team of researchers who have played a crucial role in driving the success and impact of our work. Collaboratively, we have developed and implemented the Phen2Disease methodology, leveraging the expertise and insights of individuals from various disciplines, including genomics, doctors, bioinformatics, and data analysis. Through rigorous experimentation and analysis, our researchers have meticulously examined six real data cohorts and two simulated data cohorts, comprising a total of 2051 cases. Their expertise in interpreting and evaluating the results has been instrumental in uncovering the superior performance of Phen2Disease compared to existing methods. Furthermore, our researchers have demonstrated a deep understanding of the significance of phenotype data in genomic diagnostics for rare diseases, highlighting the potential for clinical applications and contributing to the broader scientific community's knowledge base. The dedication and expertise of our research team have been integral to the development and impact of our work.

Outcomes of research leading to impact

The outcomes of our research have been instrumental in driving tangible impact across various domains. The development and implementation of the Phen2Disease methodology have significantly enhanced the prioritization of diseases and genes in genomic diagnostics for rare diseases. By leveraging bidirectional maximum matching semantic similarity between patient and disease phenotype sets, our approach has demonstrated superior performance compared to existing state-of-the-art methods, particularly in cohorts with fewer average numbers of HPO terms. This heightened accuracy and interpretability have empowered clinicians and researchers to make more informed decisions in diagnosing and treating rare diseases. The availability of ranked diseases and patient HPO terms has facilitated a deeper understanding of the underlying genomic mechanisms and has the potential to streamline the development of targeted therapies. The widespread adoption of Phen2Disease and its integration into clinical practice hold the promise of improving patient outcomes and advancing precision medicine as a whole.

Beneficiaries of the impact

The impact of our research has reached a wide range of beneficiaries across multiple sectors. Clinicians and healthcare professionals have been immediate beneficiaries, as our Phen2Disease methodology provides them with a powerful tool for prioritizing diseases and genes in genomic diagnostics for rare diseases. By accurately identifying and ranking diseases based on patient phenotypes, clinicians can make more informed decisions regarding diagnosis, treatment, and personalized care plans. Patients and their families are also beneficiaries, as they stand to gain from improved diagnostic accuracy and more targeted therapies, leading to better health outcomes and quality of life. Additionally, our research impacts the scientific community by advancing the understanding of the role of phenotype data in genomics and providing a novel approach for further research and development. The broader society benefits from our work as well, as more effective genomic diagnostics and personalized medicine contribute to overall healthcare advancements and resource optimization. Ultimately, the beneficiaries of the impact of our research span across healthcare professionals, patients, the scientific community, and society as a whole, fostering improved healthcare practices and outcomes.

Details of the impact achieved

The impact achieved through our research efforts has been substantial and far-reaching. The development and implementation of the Phen2Disease methodology have revolutionized genomic diagnostics for rare diseases. By leveraging bidirectional maximum matching semantic similarity between patient and disease phenotype sets, our approach has yielded remarkable results. Comprehensive experiments conducted on real and simulated data cohorts, encompassing a total of 2051 cases, have consistently demonstrated the superior performance of Phen2Disease compared to existing state-of-the-art methods. The enhanced accuracy and interpretability provided by Phen2Disease, such as ranked diseases and patient HPO terms, have empowered clinicians and researchers to make more precise diagnoses and treatment decisions. This, in turn, has led to improved patient outcomes, reduced diagnostic errors, and enhanced personalized care. The availability of Phen2Disease on GitHub has facilitated its widespread adoption, enabling clinicians and researchers worldwide to utilize this powerful tool for genomic diagnostics. The impact of our research extends beyond the immediate healthcare domain, as it contributes to advancing the field of precision medicine, fostering scientific advancements, and promoting resource optimization in the healthcare system. Overall, the impact achieved through our research has had a transformative effect on genomic diagnostics and holds the potential to improve the lives of countless individuals affected by rare diseases.

Impact date	2022
Category of impact	Quality of life Impact
Impact level	International