Abstract

SORL1 loss is a known risk factor for Alzheimer's disease (AD) while SORL1 overexpression is seen in various cancers. Consequently, an inverse correlation between AD and cancer has been established. As SORL1 loss in AD has a clearly defined molecular role in the formation of amyloid plaques, a similar understanding of a molecular role in cancer is missing. This project is interested in the broad molecular relationships between AD and glioma, and by exploring the molecular links between AD and cancer through SORL1, may ultimately identify new therapeutic targets. This project generated a ‘SORL1-related network’ of associated genes and proteins and built a curated SORL1-model (SKM034) of activating or inhibiting interacting targets. This was achieved through computational approach and validated in cell line data and clinical patient transcriptomic data.

The generated ‘SORL1-related network’ comprised 73 associated targets linked to SORL1. Subsequently, differential expression analysis identified 38 out of the 73 genes to be significantly different between cancer (TCGA) and non-cancer tissue (GTEx). The overall survival analysis found 92 positive association of genes in the network. In silico drug repurposing identified 122 drugs targeting nine upregulated genes in cancer in brain cancer clinical data of six CNS/Brain study groups. Boolean modelling was used to construct the SORL1 interaction network (the “SKM034 Model”) containing 34 nodes (proteins) connected by 92 interactions. Relationship analysis between SORL1 and its targets were simulated and the impact of dependency alterations assessed. In silico knockouts identified a total of 21 changes to molecular relationships following loss of SORL1. Predictions were validated in five cell line data and cancer patient RNA-seq data. Correct prediction rates for these analyses reached up to almost 60% and consistently achieved a better prediction rate than a randomised model (33.3%) in all calculation methods.

Moreover, the ‘SKM034-model’ has provided 29 potential novel predictions, which can be further explored for future research. The ‘SORL1-established network’ including all
identified genes/proteins linked to SORL1, were analysed in novel transcriptomic patient data to gain insight of differentially expressed genes in glioma with different malignancy progression versus non-cancer FFPE tissue with a total of 21 samples and subcategorised into six groups.

The findings of this project explore molecular links of SORL1 and addresses the inverse correlation between AD and cancer by computational approaches to provide novel insight into the effects of interaction partners and identify new SORL1 pathways, specifically the role of zinc in AD and cancer. Moreover, establishing PLAUR as a potential biomarker for AD and glioma. The SKM034 model of SORL1 applied a novel approach to SORL1 research and further established a molecular link between AD and glioma.