06 November 2021

Summary: Machine learning and artificial intelligence approaches for PH: From screening to novel drug targets

October 25, 2019 (73).jpg

Event summary

‘Machine learning and artificial intelligence approaches for PH: From Screening to novel drug targets’ webinar focused on novel applications of machine learning and artificial intelligence that are pushing pulmonary hypertension (PH) research forward. The average PH journey from symptoms to diagnosis takes about 2 years due to misdiagnoses, late treatment of advanced disease, and finding the best treatment for nonspecific symptoms. Artificial Intelligence (AI) approaches are being used to make more informed decisions for drug development, patient classification or stratification using -omics, imaging, and clinical features. Large patient healthcare datasets are being used to identify footprints of idiopathic disease that can potentially improve the diagnostic time and understand the heterogenous nature of PH.

Dr Dennis Wang (University of Sheffield, UK) provided the practical challenges and applications of mainly supervised machine learning methods in translational medicine1. He discussed Alphafold2, an AI to predict protein structure with high accuracy with the potential to speed up target identification, binding, and interactions with the folds in proteins. AI also presents blackbox challenges that may reflect relationships that may not make biophysical sense.

Dr Joseph Loscalzo (Harvard University, USA) focused on the use of network analysis of protein-protein interactomes to identify drug targets, and to repurpose drugs for other diseases.  Network analysis of interactomes has been used to predict drug responses in SARS-CoV-23 and drug interactions in PAH therapies. In vitro testing is still essential, as new toxicities may arise when repurposing drugs. The panel agreed that modern computational and molecular interaction network-based drug development strategies offer clear advantages over conventional drug development.

Dr David Kiely (University of Sheffield, UK) looked at the use of AI to improve the diagnostic performance of tests to identify PH. AI can be leveraged to produce faster diagnostic times, provide signatures of PH that may not be visible to the naked eye, predict risk of PH-related hospitalization, or provide insight into the cause of PH. Many screening modalities such as chest x-ray, ECG, blood biomarkers, and echocardiography are available. Dr Kiely also discussed the SPHinX project4,5, that applied AI to large datasets from real-world healthcare data resources to determine iPAH diagnosis. Results support the use of large patient data in diagnostic care but the challenge is maximizing the potential of AI for improving outcomes in PAH. Polls expressed that many think that AI approaches will reduce the time to a diagnosis of PAH within the next 5 years.

As part of the best abstract discussions, Sokratis Kariotis (University of Sheffield, UK) looked at the heterogenous population in iPAH, and if patient subgroups can be defined based on RNA-seq transcriptomic profiles and clinical data profiles. The study identified 5 distinct iPAH subgroups, and immune genes were found to be reflective of severity and survival. Yajing Ji (Michigan State University, USA) investigated the identification of signaling pathways that are altered in PAH. Unsupervised hierarchical clustering was used to identify molecular PAH subgroups based on the patient classification and the predictive genesets. This combined meta-analysis identified three subclasses of low, medium and high levels of inflammatory pathways.

Summary by Alexandra Janowski, The Ohio State University, USA

Please note that the recording of this event is only available to members. All PVRI2021 webinars will be made public up to 12 weeks after the event date. Please sign up for news on the sidebar of this page to be alerted when this recording will be made available to non-members.

References

  1. Toh, T. S., Dondelinger, F. & Wang, D. Looking beyond the hype: Applied AI and machine learning in translational medicine. EBioMedicine 47, 607–615 (2019).
  2. Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
  3. Morselli Gysi, D. et al. Network medicine framework for identifying drug-repurposing opportunities for COVID-19. Proc Natl Acad Sci U S A 118, e2025581118 (2021).
  4. Bergemann, R. et al. High levels of healthcare utilization prior to diagnosis in idiopathic pulmonary arterial hypertension support the feasibility of an early diagnosis algorithm: the SPHInX project. Pulm Circ 8, (2018).
  5. Kiely, D. G. et al. Utilising artificial intelligence to determine patients at risk of a rare disease: idiopathic pulmonary arterial hypertension. Pulm Circ 9, 2045894019890549 (2019).

Topics

Pulmonary Arterial Hypertension
Pulmonary Hypertension
Transcriptomics
Translational Research
Phenotype
Diagnosis


Comments (0)

Our research platform is the world.

Through worldwide collaboration, we can begin to answer the question of a global disease.

Join the PVRI
standard-example-image.jpg