Y CAO1, A LEE1, S SHAZANFAR1, S ALEXANDER2, J CRAIG2, G WONG2, J YANG3
1The University of Sydney, Camperdown, Australia, 2Centre for Kidney Research, Kids Research Institute, Westmead Hospital; College of Medicine and Public Health, Flinders University, , Australia, 3Charles Perkins Centre, The University of Sydney; School of Mathematics and Statistics, University of Sydney, Camperdown, Australia
Aim: To develop a validated set of models that identify transcriptomics-based biomarkers for predicting acute rejection after kidney transplantation.
Background: Identification of biomarkers using transcriptional profiling may allow prediction of recipients who are at risk of developing allograft injury such as acute rejection after transplantation.
Methods: Comprehensive search using Gene Expression Omnibus was conducted for publicly available datasets that contained gene expression and acute rejection data after kidney transplantation. With each selected dataset, a set of biomarkers for acute rejection was derived using five algorithms including, but not limited to XgBoost, Lasso and Random Forest. Biomarkers for each dataset are then cross-validated on the remaining independent datasets.
Results: A total of six datasets (n = 738) were included in the meta-analyses. For each dataset, a set of genes was identified for prediction of acute rejection patients within its own dataset, with optimal performance of the modeling achieved using XgBoost classifier (mean accuracy of six datasets > 84%, SD = 0.053). A set of 129 gene loci that included CXCL6, CXCL11 and OLFM4 achieved accuracy over 70% for prediction of acute rejection patients across three of the six datasets. These 129 gene loci (with 175 gene-gene interaction, based on String) were significantly enriched for genes in both the innate and cognate immune system including a significant number of molecules involved in immune trafficking such as adhesion molecules, chemokines and chemokine receptors and a number of as yet uncharacterised but intriguing zinc finger transcription factors.
Conclusions: A set of 129 gene loci with many found in immune-driven molecular pathways including adhesion, trafficking and activation achieved satisfactory performance in predicting acute rejection after kidney transplantation.
I am currently in my Honours year of the Bachelor of Science (Advanced) degree at the University of Sydney. I double major in Molecular Biology and Genetics and Computer Science. I enjoy working with biological/clinical data, and is particularly intrigued by how programming can be utilised to discover patterns and drive solutions.