ChIPXpress

ChIPXpress integrates ChIPx data with publicly available gene expression datasets to prioritize functional transcription factor (TF) target genes that respond transcriptionally to TF perturbation.

Key Features:

• Data integration: Integrates ChIPx experiments (ChIP-seq and ChIP-chip) with publicly available gene expression datasets (PED) to inform target prioritization.
• Truncated absolute correlation: Employs a truncated absolute correlation measure to capture regulatory relationships between TFs and target genes from heterogeneous PED.
• Combined scoring and ranking: Produces ranked lists of TF-bound genes by combining ChIPx-derived signals with PED-derived correlation information.
• Robustness to peak caller: Demonstrates improved ranking accuracy irrespective of the peak calling algorithm used on ChIPx data.
• Addresses missing perturbation data: Designed to improve prioritization when gene expression data from TF perturbation experiments are absent in studies (noted in ~40% of cases).
• Implementation: Implemented as an R/Bioconductor package.
• Evaluation: Evaluated on 10 diverse ChIPx datasets from mouse and human studies.

Scientific Applications:

• Prioritization of functional TF targets: Ranks TF-bound genes to enrich for targets that are transcriptionally responsive to TF perturbation.
• Experimental candidate selection: Helps select top-ranked candidate targets for follow-up perturbation or validation experiments.
• Augmenting ChIPx analyses: Complements ChIPx-only analyses by leveraging PED to improve prediction reliability in functional genomics studies.
• Cross-study leverage of expression data: Enables use of heterogeneous public gene expression datasets to inform TF-target relationships across studies.

Methodology:

Integrates ChIPx data with publicly available gene expression datasets and applies a truncated absolute correlation measure to score regulatory relationships, combines these scores with ChIPx signal to rank targets, and was evaluated on 10 mouse and human ChIPx datasets across different peak calling algorithms.