deepTMHMM Topology Variation — Interesting Genes

Genes with transmembrane topology variation across haplotypes (filtered from 2,355 genes with ≥2 distinct topologies in JoGo)

Help — deepTMHMM Topology Variation Viewer

Overview

This page lists 2,355 genes from the JoGo haplotype-resolved proteome where the predicted transmembrane topology varies across haplotypes (a-level). Predictions were generated by deepTMHMM, a deep-learning method for transmembrane topology prediction. Only genes with ≥2 distinct topologies are shown.

Data Source

Haplotype sequences: JoGo a-level haplotypes (174,376 sequences from 19,193 gene regions)
Prediction tool: deepTMHMM (DTU Biolib)
Gene annotations: MANE Select v1.2 (GRCh38), 19,316 genes
Database: deepTMHMM.db — 597,769 topology segment rows across 19,193 gene regions

Topology States

State	Color	Description
TMhelix	Red	Transmembrane helix — alpha-helical segment spanning the membrane
signal	Orange	Signal peptide — N-terminal targeting sequence cleaved after translocation
inside	Blue	Cytoplasmic (inside) region
outside	Green	Extracellular (outside) / lumenal region
Beta sheet	Teal	Transmembrane beta-barrel sheet (outer membrane proteins)
periplasm	Purple	Periplasmic region (rare in human proteins)

Table Columns

Column	Description
Interest	Star rating (1–5) based on the interest score formula below
Gene	MANE gene symbol. Click to open the per-gene topology viewer page
Region	Genomic region name (`GENE_chrN_start_end`). Click to open JoGo browser
Prot Len	Maximum protein length (aa) across haplotypes
Haplotypes	Number of distinct a-level haplotypes for this gene
Topologies	Number of distinct topology architectures across haplotypes
TM Range	min–max number of TM helices across haplotypes (e.g., `0-7` means some haplotypes have 0 TM helices and some have 7)
Total Diffs	Sum of all topology differences vs reference haplotype (a0001)
Diff Breakdown	Stacked bar showing proportion of each diff type (see colors below)
Counts	Badges showing count per diff type
Haps w/ TM	Number of haplotypes that have at least one TM helix
Haps w/ SP	Number of haplotypes that have a signal peptide

Difference Types

Type	Badge	Description
TM count change	N TM	Haplotype has a different number of TM helices than the reference (a0001)
Signal change	N SP	Haplotype gained or lost a signal peptide vs reference
Boundary shift	N shift	Same TM/signal count as reference but different segment boundaries

Interest Score

Genes are ranked by a composite interest score:

score = n_topologies × 5.0 + n_total_diffs × 0.1 + tm_range × 3.0 + tm_change_bonus (2.0 × min(n_tm_change, 10)) + signal_change_bonus (4.0 if any)

Higher scores indicate more biologically interesting topology variation. Genes with large TM count ranges, many distinct topologies, and signal peptide changes rank highest.

Filters

Filter	Description
Search	Filter by gene name (case-insensitive substring match)
Min Topologies	Only show genes with at least N distinct topology architectures
Min Diffs	Only show genes with at least N total differences vs reference
Has TM count change	At least one haplotype has a different TM helix count than reference
Has signal change	At least one haplotype differs in signal peptide presence
TM range ≥ 3	The difference between max and min TM helix count is ≥ 3
Some haps 0 TM, some >0	Some haplotypes have no TM helices while others do — most dramatic variation

Per-Gene Viewer

Clicking a gene name opens the detailed topology viewer page (REGION_tmhmm_viewer.html) which shows:

Gene Information — MANE gene metadata, TM helix count, topology string
Topology Diagram — Interactive SVG showing all segments per haplotype with hover tooltips
Topology Differences — Table of per-haplotype differences vs reference
Haplotype Details — Collapsible per-haplotype segment tables
Summary Statistics — TM helix distribution and diff breakdown

Methods

The pipeline consists of the following steps:

Extract amino acid sequences from JoGo haplotype TSV (174,376 sequences, filtering 518 zero-length entries)
Split into 174 chunks of 1,000 sequences each
Run deepTMHMM via Singularity container on SLURM cluster (60 cores, 128 GB per job)
Parse GFF3 output and reformat into TSV
Build SQLite database with deepTMHMM predictions + MANE gene annotations
Generate per-gene HTML viewer pages (19,193 files) and this summary page

deepTMHMM Topology Variation — Interesting Genes ?