Deep graph learning approaches to personalized medicine

Funder

Organizational Unit

Publications

Biomedical knowledge graph embeddings for personalized medicine: Predicting disease‐gene associations

Publication . Vilela, Joana; Asif, Muhammad; Marques, Ana Rita; Santos, João Xavier; Rasga, Célia; Vicente, Astrid; Martiniano, Hugo

Personalized medicine is a concept that has been subject of increasing interest in medical research and practice in the last few years. However, significant challenges stand in the way of practical implementations, namely in regard to extracting clinically valuable insights from the vast amount of biomedical knowledge generated in the last few years. Here, we describe an approach that uses Knowledge Graph Embedding (KGE) methods on a biomedical Knowledge Graph (KG) as a path to reasoning over the wealth of information stored in publicly accessible databases. We built a Knowledge Graph using data from DisGeNET and GO, containing relationships between genes, diseases and other biological entities. The KG contains 93,657 nodes of 5 types and 1,705,585 relationships of 59 types. We applied KGE methods to this KG, obtaining an excellent performance in predicting gene-disease associations (MR 0.13, MRR 0.96, HITS@1 0.93, HITS@3 0.99, and HITS@10 0.99). The optimal hyperparameter set was used to predict all possible novel gene-disease associations. An in-depth analysis of novel gene-disease predictions for disease terms related to Autism Spectrum Disorder (ASD) shows that this approach produces predictions consistent with known candidate genes and biological pathways and yields relevant insights into the biology of this paradigmatic complex disorder.

2022-11-20Journal article

Open access

Publication . Vilela, Joana; Martiniano, Hugo; Marques, Ana Rita; Santos, João Xavier; Rasga, Célia; Oliveira, Guiomar; Vicente, Astrid Moura

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder with heterogeneous clinical presentation, variable severity, and multiple comorbidities. A complex underlying genetic architecture matches the clinical heterogeneity, and evidence indicates that several co-occurring brain disorders share a genetic component with ASD. In this study, we established a genetic similarity disease network approach to explore the shared genetics between ASD and frequent comorbid brain diseases (and subtypes), namely Intellectual Disability, Attention-Deficit/Hyperactivity Disorder, and Epilepsy, as well as other rarely co-occurring neuropsychiatric conditions in the Schizophrenia and Bipolar Disease spectrum. Using sets of disease-associated genes curated by the DisGeNET database, disease genetic similarity was estimated from the Jaccard coefficient between disease pairs, and the Leiden detection algorithm was used to identify network disease communities and define shared biological pathways. We identified a heterogeneous brain disease community that is genetically more similar to ASD, and that includes Epilepsy, Bipolar Disorder, Attention-Deficit/Hyperactivity Disorder combined type, and some disorders in the Schizophrenia Spectrum. To identify loss-of-function rare de novo variants within shared genes underlying the disease communities, we analyzed a large ASD whole-genome sequencing dataset, showing that ASD shares genes with multiple brain disorders from other, less genetically similar, communities. Some genes (e.g., SHANK3, ASH1L, SCN2A, CHD2, and MECP2) were previously implicated in ASD and these disorders. This approach enabled further clarification of genetic sharing between ASD and brain disorders, with a finer granularity in disease classification and multi-level evidence from DisGeNET. Understanding genetic sharing across disorders has important implications for disease nosology, pathophysiology, and personalized treatment.

2022-08-18Journal article

Open access

Publication . Vilela, Joana; Martiniano, Hugo; Marques, Ana Rita; Santos, João; Rasga, Célia; Oliveira, Guiomar; Vicente, Astrid Moura

Background/Objectives: Autism Spectrum Disorder (ASD) is a clinically heterogeneous neurodevelopmental disorder , with variable severity and multiple comorbidities. Given the previous evidence of genetic overlap between ASD and several comorbid brain disorders, we sought to explore the genetic similarity with ASD across a range of brain disorders, using a genetic similarity disease network approach. Methods: We developed a genetic similarity disease network between ASD and Intellectual Disability, Attention-Deficit/Hyperactivity Disorder, Epilepsy, Schizophrenia and Bipolar Disease spectrum. Using gene-disease associations from the DisGeNET database, genetic similarities were estimated from the Jaccard coefficient between disease pairs. The Leiden algorithm identified network disease communities and shared biological pathways. Loss-of-function (LoF) rare de novo variants within shared genes underlying the disease communities were identified using the MSSNG whole-genome sequencing dataset. Results: We identified three disease communities. ASD is included in a heterogeneous community with Epilepsy, Bipolar Disorder, Attention-Deficit/Hyperactivity Disorder combined type, and some disorders in the Schizophrenia Spectrum. ASD and Intellectual Disability are in separate communities. The genes SHANK3, ASH1L, SCN2A, and CHD2, which are candidate genes for diseases in all communities, have a higher number of de novo rare LoF SNVs in ASD subjects. Conclusion: This approach enabled further clarification of genetic sharing between ASD and comorbid brain disorders, as we took advantage from a finer granularity in disease classification and multi-level evidence from DisGeNET, with important implications for disease nosology, pathophysiology, and personalized treatment.

2023-06-10Conference object

Open access