Discovering causal pathways linking genomic events to transcriptional states using Tied Diffusion Through Interacting Events (TieDIE)

Abstract

Motivation: Identifying the cellular wiring that connects genomic perturbations to transcriptional changes in cancer is essential to gain a mechanistic understanding of disease initiation, progression and ultimately to predict drug response. We have developed a method called Tied Diffusion Through Interacting Events (TieDIE) that uses a network diffusion approach to connect genomic perturbations to gene expression changes characteristic of cancer subtypes. The method computes a subnetwork of protein-protein interactions, predicted transcription factor-to-target connections and curated interactions from literature that connects genomic and transcriptomic perturbations.

Results: Application of TieDIE to The Cancer Genome Atlas and a breast cancer cell line dataset identified key signaling pathways, with examples impinging on MYC activity. Interlinking genes are predicted to correspond to essential components of cancer signaling and may provide a mechanistic explanation of tumor character and suggest subtype-specific drug targets.

Availability: Software is available from the Stuart lab's wiki: https://sysbiowiki.soe.ucsc.edu/tiedie.

Contact: jstuart@ucsc.edu.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Schematic of TieDIE. Relevant genes from two distinct sets are shown as nodes colored by dyes diffusing on a network from a source set (e.g. mutated genes; red nodes) and target set (e.g. TFs; blue nodes). ‘Linker’ genes (purple nodes) residing between the source and target sets are revealed through a diffusion process evolved over time; two time slices are shown as stacked layers of the same network

Fig. 2.

Frequency of a discovered core and off-core genes in single-source and tied-diffusion in a simulated network. (A) Single-source diffusion over the synthetic network. Darker colors indicate genes in a larger fraction of network solutions in repeated simulated trials at a fixed recall of 4 of 6 signaling genes. (B) The corresponding tied-diffusion frequencies at identical recall and test conditions

Fig. 3.

Precision of single-source (blue points) and tied-diffusion (orange points) with different relevance scores for identifying pathways in a breast cancer. Any paths containing even a single randomly injected ‘decoy’ link were considered false positives. Recall measures the number of logically consistent paths (X-axis; see Methods) out of the total possible; precision measures the number of such consistent paths in the total number returned. Relevance scores tested are heat diffusion (circles), personalized PageRank (triangles) and SPIA (green circles). For comparison, included are all-pairs shortest paths (APSP; blue circle) and prize-collecting Steiner trees (PCST; red dot). Randomly generated networks of various sizes were obtained to estimate the background distribution (gray dots). Different levels of precision and recall were obtained by varying algorithm parameters (e.g. the α parameter for single and tied diffusion; Supplementary Methods S1.5)

Fig. 4.

Tied-diffusion result for luminal A versus basal breast cancer subtypes. The inner coloring of the rings represents the differential expression in luminal A as compared with basal samples. The outer ring represents differential frequency of genomic perturbations in luminal samples as compared with basal samples: differential mutation (upper right), amplification (lower right), deletion (lower left) and DNA-methylated CpG islands near the promoter (upper left)

Fig. 5.

Luminal A sample TCGA-BH-A0BR specific network reveals basal-like molecular behavior. The network connects genomic perturbations in the sample, red or blue rings around nodes, to transcriptional changes in the same sample, inner node coloring. Red and blue colors indicate higher and lower cohort mutation rates in luminal A samples as compared with basal, outer ring; overall cohort differential expression of luminal A compared with basal samples, second ring; individual sample expression, inner circle. Transcriptional interactions, solid lines; post-transcriptional interactions, dashed. Activating interactions, arrow at the target node; inactivating, flat bars