Here is another Breakdown.
Sure — here's the step-by-step breakdown of the molecular mechanism types INFERNO actually probes, using the data sources you listed.
🧬 The Central Mechanistic Question
A GWAS says SNP rs12345 is associated with Crohn's disease (
p
=
1
×
10
−
12
p=1×10−12). But:
•rs12345 is noncoding — it doesn't change an amino acid
•It's in LD with 47 other SNPs — which one is causal?
•Even if you identify the causal SNP, what does it actually do?
INFERNO answers this by testing each candidate SNP against specific molecular mechanism categories:
🔍 Mechanism 1: Transcription Factor Binding Disruption
What it is: A SNP falls inside a TF binding motif, altering the binding affinity of a transcription factor.
How INFERNO detects it:
•FANTOM5 provides CAGE-defined enhancer and promoter boundaries
•ENCODE/Roadmap TF ChIP-seq peaks show where specific TFs actually bind
•INFERNO overlaps the candidate SNP coordinates with these TF binding regions
•If the SNP alters a nucleotide within a known position weight matrix (PWM) for a TF, that's a mechanistic hit
Molecular consequence: TF X binds less (or more) tightly → nearby gene transcription goes up or down.
🧬 Mechanism 2: Enhancer/Promoter Activity Alteration
What it is: The SNP falls in an enhancer or promoter region and alters its regulatory activity.
How INFERNO detects it:
•FANTOM5 CAGE data defines active enhancers and promoters across hundreds of cell types and tissues — since CAGE captures the 5' end of transcribed RNAs, it marks both mRNA transcription start sites (promoters) and bidirectional enhancer RNAs (active enhancers)
•ChromHMM 15-state chromatin models (trained on histone marks like H3K27ac, H3K4me1, H3K4me3, H3K36me3, etc.) classify every genomic region into states like "Active Enhancer," "Poised Promoter," "Transcribed," "Repressed Polycomb," etc.
•INFERNO checks whether the SNP falls in a ChromHMM state that differs between relevant cell types
Molecular consequence: The enhancer becomes stronger/weaker → target gene expression changes in a tissue-specific manner.
🧫 Mechanism 3: Tissue/Cell-Type-Specific Regulatory Effects
What it is: The causal variant only exerts its effect in specific tissues or cell types.
How INFERNO detects it:
•GTEx eQTL data across ~49 tissues lets INFERNO check whether a GWAS SNP is an eQTL only in disease-relevant tissues
•FANTOM5 enhancer annotations are tissue-specific — an enhancer active in pancreatic islets but silent in brain
•ChromHMM state calls are also cell-type-specific — the same genomic coordinate can be "Active Enhancer" in CD4 T cells but "Quiescent" in hepatocytes
Molecular consequence: Explains why a variant causes type 2 diabetes but not neurological disease — the regulatory element it disrupts is only active in pancreatic beta cells…