Molecular Inverse Comorbidity between Alzheimer's Disease and Lung Cancer: New Insights from Matrix Factorization

Abstract

Matrix factorization (MF) is an established paradigm for large-scale biological data analysis with tremendous potential in computational biology. Here, we challenge MF in depicting the molecular bases of epidemiologically described disease-disease (DD) relationships. As a use case, we focus on the inverse comorbidity association between Alzheimer's disease (AD) and lung cancer (LC), described as a lower than expected probability of developing LC in AD patients. To this day, the molecular mechanisms underlying DD relationships remain poorly explained and their better characterization might offer unprecedented clinical opportunities. To this goal, we extend our previously designed MF-based framework for the molecular characterization of DD relationships. Considering AD-LC inverse comorbidity as a case study, we highlight multiple molecular mechanisms, among which we confirm the involvement of processes related to the immune system and mitochondrial metabolism. We then distinguish mechanisms specific to LC from those shared with other cancers through a pan-cancer analysis. Additionally, new candidate molecular players, such as estrogen receptor (ER), cadherin 1 (CDH1) and histone deacetylase (HDAC), are pinpointed as factors that might underlie the inverse relationship, opening the way to new investigations. Finally, some lung cancer subtype-specific factors are also detected, also suggesting the existence of heterogeneity across patients in the context of inverse comorbidity.

Keywords: Alzheimer’s disease; inverse comorbidity; lung cancer; matrix factorization; networks; transcriptome.

Figure 1

Schematic view of the framework and the novelties introduced with respect to [12]. RBH = reciprocal best hit. MF = matrix factorization. AD = Alzheimer’s disease. LC = lung cancer.

Figure 2

Schematic representation of the metagene orientation procedures: Long-tail-pointing (A) and disease-pointing (B). In (A), each couple (metagene, metasample) is oriented based on the distribution of the metagene weights. Two possible scenarios may thus be verified depending on the distribution of the weights constituting the metagene: if the long tail of the distribution is already in the positive side, no operation is performed; if instead the long tail is pointing in the negative direction, the orientation of the two vectors is inverted. In (B), each couple (metagene, metasample) is oriented based on the distribution of the metasample weights. Two possible scenarios may thus be verified depending on the case vs. control distribution of the weights constituting the metasample: if case samples have higher weights with respect to control samples, no operation is performed; in the opposite case, the two vectors are inverted.

Figure 3

“Long-tail-pointing” (red) vs. “disease-pointing” (blue) orientation of the stabilized independent component analysis (sICA) factors. (A) The two methods of factor orientation are compared based on the correlation of the obtained metagenes with the case vs. control genes’ fold change. (B) The two methods are compared based on the number of links present in their RBH network. Total RBHs (RBH), positive RBHs (+RBH), negative RBHs (–RBH). (C–E) The two methods are compared based on the structure of their –RBH AD/LC subnetwork, relevant for the study of inverse comorbidity. In (C), the number of nodes and links in the subnetwork are compared. Number of nodes (# Nodes), negative RBHs connecting an AD component with a LC component (–RBH) and negative RBHs connecting an AD component with a LC component that are associated to nodes with significant differential behaviour (Wilcoxon p-value < 0.05) between case and control (significant –RBH). In (D), the clustering coefficient and modularity of the subnetwork are considered. In (E), the number of communities and their average size in the subnetwork is taken into account.

Figure 4

–RBH AD/LC subnetwork with biological annotations. Each node in the network corresponds to a metagene; the list of metagenes associated to each community ID is reported in Supplementary Table S3. Colours are linked to the diseases: red for AD and blue for LC. In AD, datasets obtained from the same region of the brain are denoted with different shades of red (normal and light red). The nodes are organized into communities. Each community is denoted with a number corresponding to its ID and the main biological annotation associated to them (see Supplementary Table S2 for an extensive report). HDAC = histone deacetylase.