# agreed_h.tsv.gz

Preliminary. For the CATH v4.5.0 prerelease.

## What this data represents

TED v6 domains for which Contrasted and Foldseek assign the same CATH v4.4 superfamily.

Gzipped tab-separated text. Header on line 1. One domain per row.

## How the data was generated

Each TED v6 domain was scored with Contrasted 1-NN and with a Foldseek rank-1 3Di+AA hit. A row is included when both methods pass the superfamily (H) gate and assign the same superfamily.

Contrasted passes when 1-cosine <= 0.25, the domain-length ratio min(Lq, Lt)/max(Lq, Lt) >= 0.60, and the neighbour has a CATH superfamily label.

Foldseek hits in the input already pass E <= 0.05, min(qTM, tTM) >= 0.50, and min(qcov, tcov) >= 0.60. Foldseek assigns an H when E <= 0.01 and the hit has a superfamily label.

296,093,143 TED v6 domains were scored. Domains where the two methods do not agree on an H are in `held.tsv.gz` or `t_seeds.tsv.gz`. 116,894,232 domains where both methods abstain are omitted.

## Columns

| Column | Meaning |
|---|---|
| query_id | TED v6 domain |
| h | Superfamily assigned by both methods |
| contrasted_neighbor | Contrasted rank-1 neighbour |
| distance | Contrasted 1-cosine |
| length_ratio | min(Lq, Lt) / max(Lq, Lt) |
| foldseek_neighbor | Foldseek rank-1 hit |
| evalue | Foldseek E-value |
| qtmscore | Foldseek query TM |
| ttmscore | Foldseek target TM |
| qcov | Foldseek query coverage |
| tcov | Foldseek target coverage |

## Shape

143,732,518 data rows.