Explainable artificial intelligence in high-throughput drug repositioning for subgroup stratifications with interventionable potential

Automated identification of patient subgroups: A case-study on mortality of COVID-19 patients admitted to the ICU

BACKGROUND

- Subgroup discovery (SGD) is the automated splitting of the data into complex subgroups. Various SGD methods have been applied to the medical domain, but none have been extensively evaluated. We assess the numerical and clinical quality of SGD methods.

METHOD

- We applied the improved Subgroup Set Discovery (SSD++), Patient Rule Induction Method (PRIM) and APRIORI - Subgroup Discovery (APRIORI-SD) algorithms to obtain patient subgroups on observational data of 14,548 COVID-19 patients admitted to 73 Dutch intensive care units. Hospital mortality was the clinical outcome. Numerical significance of the subgroups was assessed with information-theoretic measures. Clinical significance of the subgroups was assessed by comparing variable importance on population and subgroup levels and by expert evaluation.

RESULTS

- The tested algorithms varied widely in the total number of discovered subgroups (5-62), the number of selected variables, and the predictive value of the subgroups. Qualitative assessment showed that the found subgroups make clinical sense. SSD++ found most subgroups (n = 62), which added predictive value and generally showed high potential for clinical use. APRIORI-SD and PRIM found fewer subgroups (n = 5 and 6), which did not add predictive value and were clinically less relevant.

CONCLUSION

- Automated SGD methods find clinical subgroups that are relevant when assessed quantitatively (yield added predictive value) and qualitatively (intensivists consider the subgroups significant). Different methods yield different subgroups with varying degrees of predictive performance and clinical quality. External validation is needed to generalize the results to other populations and future research should explore which algorithm performs best in other settings.

Network neighborhood operates as a drug repositioning method for cancer treatment

Computational drug repositioning approaches are important, as they cost less compared to the traditional drug development processes. This study proposes a novel network-based drug repositioning approach, which computes similarities between disease-causing genes and drug-affected genes in a network topology to suggest candidate drugs with highest similarity scores. This new method aims to identify better treatment options by integrating systems biology approaches. It uses a protein-protein interaction network that is the main topology to compute a similarity score between candidate drugs and disease-causing genes. The disease-causing genes were mapped on this network structure. Transcriptome profiles of drug candidates were taken from the LINCS project and mapped individually on the network structure. The similarity of these two networks was calculated by different network neighborhood metrics, including Adamic-Adar, PageRank and neighborhood scoring. The proposed approach identifies the best candidates by choosing the drugs with significant similarity scores. The method was experimented on melanoma, colorectal, and prostate cancers. Several candidate drugs were predicted by applying AUC values of 0.6 or higher. Some of the predictions were approved by clinical phase trials or other in-vivo studies found in literature. The proposed drug repositioning approach would suggest better treatment options with integration of functional information between genes and transcriptome level effects of drug perturbations and diseases.

Overcoming cohort heterogeneity for the prediction of subclinical cardiovascular disease risk

Cardiovascular disease remains a leading cause of mortality with an estimated half a billion people affected in 2019. However, detecting signals between specific pathophysiology and coronary plaque phenotypes using complex multi-omic discovery datasets remains challenging due to the diversity of individuals and their risk factors. Given the complex cohort heterogeneity present in those with coronary artery disease (CAD), we illustrate several different methods, both knowledge-guided and data-driven approaches, for identifying subcohorts of individuals with subclinical CAD and distinct metabolomic signatures. We then demonstrate that utilizing these subcohorts can improve the prediction of subclinical CAD and can facilitate the discovery of novel biomarkers of subclinical disease. Analyses acknowledging cohort heterogeneity through identifying and utilizing these subcohorts may be able to advance our understanding of CVD and provide more effective preventative treatments to reduce the burden of this disease in individuals and in society as a whole.

Trends and opportunities in computable clinical phenotyping: A scoping review

Identifying patient cohorts meeting the criteria of specific phenotypes is essential in biomedicine and particularly timely in precision medicine. Many research groups deliver pipelines that automatically retrieve and analyze data elements from one or more sources to automate this task and deliver high-performing computable phenotypes. We applied a systematic approach based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines to conduct a thorough scoping review on computable clinical phenotyping. Five databases were searched using a query that combined the concepts of automation, clinical context, and phenotyping. Subsequently, four reviewers screened 7960 records (after removing over 4000 duplicates) and selected 139 that satisfied the inclusion criteria. This dataset was analyzed to extract information on target use cases, data-related topics, phenotyping methodologies, evaluation strategies, and portability of developed solutions. Most studies supported patient cohort selection without discussing the application to specific use cases, such as precision medicine. Electronic Health Records were the primary source in 87.1 % (N = 121) of all studies, and International Classification of Diseases codes were heavily used in 55.4 % (N = 77) of all studies, however, only 25.9 % (N = 36) of the records described compliance with a common data model. In terms of the presented methods, traditional Machine Learning (ML) was the dominant method, often combined with natural language processing and other approaches, while external validation and portability of computable phenotypes were pursued in many cases. These findings revealed that defining target use cases precisely, moving away from sole ML strategies, and evaluating the proposed solutions in the real setting are essential opportunities for future work. There is also momentum and an emerging need for computable phenotyping to support clinical and epidemiological research and precision medicine.

Artificial intelligence in pharmacology research and practice

In recent years, the use of artificial intelligence (AI) in health care has risen steadily, including a wide range of applications in the field of pharmacology. AI is now used throughout the entire continuum of pharmacology research and clinical practice and from early drug discovery to real-world datamining. The types of AI models used range from unsupervised clustering of drugs or patients aimed at identifying potential drug compounds or suitable patient populations, to supervised machine learning approaches to improve therapeutic drug monitoring. Additionally, natural language processing is increasingly used to mine electronic health records to obtain real-world data. In this mini-review, we discuss the basics of AI followed by an outline of its application in pharmacology research and clinical practice.

Measurement of lipid flux to advance translational research: evolution of classic methods to the future of precision health

Over the past 70 years, the study of lipid metabolism has led to important discoveries in identifying the underlying mechanisms of chronic diseases. Advances in the use of stable isotopes and mass spectrometry in humans have expanded our knowledge of target molecules that contribute to pathologies and lipid metabolic pathways. These advances have been leveraged within two research paths, leading to the ability (1) to quantitate lipid flux to understand the fundamentals of human physiology and pathology and (2) to perform untargeted analyses of human blood and tissues derived from a single timepoint to identify lipidomic patterns that predict disease. This review describes the physiological and analytical parameters that influence these measurements and how these issues will propel the coming together of the two fields of metabolic tracing and lipidomics. The potential of data science to advance these fields is also discussed. Future developments are needed to increase the precision of lipid measurements in human samples, leading to discoveries in how individuals vary in their production, storage, and use of lipids. New techniques are critical to support clinical strategies to prevent disease and to identify mechanisms by which treatments confer health benefits with the overall goal of reducing the burden of human disease.

Personalized tracking of how lipid (fat) metabolism changes over time could lead to improvements in the diagnosis and treatment of several diseases. Elizabeth Parks and colleagues from the University of Missouri, Columbia, USA, discuss the ways in which researchers use stable isotope labeling to monitor the kinetics of fatty acids and other lipids in the body. Usually, lipid quantities are measured only at a single timepoint, however the tracking of lipid turnover over time provides further diagnostic information. Aided by new techniques such as high-throughput mass spectrometry and machine learning, researchers are now able to continuously map total lipid contents in individual patients. The transition of measurements of lipid flux from the research laboratory to the doctor’s office will likely play a role in a new era of precision medicine.

Drug Repositioning and Subgroup Discovery for Precision Medicine Implementation in Triple Negative Breast Cancer

Simple Summary

The heterogeneity of complicated diseases like cancer negatively affects patients’ responses to treatment. Finding homogeneous subgroups of patients within the cancer population and finding the appropriate treatment for each subgroup will improve patients’ survival. In this study, we focus on triple-negative breast cancer (TNBC), where approximately 80% of patients do not entirely respond to chemotherapy. Our aim is to find subgroups of TNBC patients and identify drugs that have the potential to tailor treatments for each group through drug repositioning. After applying our method to TNBC, we found that different targeted mechanisms were suggested for different groups of patients. Our findings could help the research community to gain a better understanding of different subgroups within the TNBC population and can help the drugs to be repurposed with explainable results regarding the targeted mechanism.

Abstract

Breast cancer (BC) is the leading cause of death among female patients with cancer. Patients with triple-negative breast cancer (TNBC) have the lowest survival rate. TNBC has substantial heterogeneity within the BC population. This study utilized our novel patient stratification and drug repositioning method to find subgroups of BC patients that share common genetic profiles and that may respond similarly to the recommended drugs. After further examination of the discovered patient subgroups, we identified five homogeneous druggable TNBC subgroups. A drug repositioning algorithm was then applied to find the drugs with a high potential for each subgroup. Most of the top drugs for these subgroups were chemotherapy used for various types of cancer, including BC. After analyzing the biological mechanisms targeted by these drugs, ferroptosis was the common cell death mechanism induced by the top drugs in the subgroups with neoplasm subdivision and race as clinical variables. In contrast, the antioxidative effect on cancer cells was the common targeted mechanism in the subgroup of patients with an age less than 50. Literature reviews were used to validate our findings, which could provide invaluable insights to streamline the drug repositioning process and could be further studied in a wet lab setting and in clinical trials.