---
title: Humatch inputs and parameters
description: Up to 50 sequences, the chain_type that picks the heavy or light model — mislabel it and the score is meaningless; the threshold moves only the human call.
tags:
  - humatch
  - inputs
---

You submit sequences in one of two shapes, and the shape decides which classifiers run. Loose chains go in `sequences` and are scored by the per-chain heavy or light CNN. Heavy/light pairs go in `chain_pairs` and get both — each chain's own per-chain score **and** Humatch's paired classifier, which exists only for pairs and cannot be approximated from one chain alone.

The chain type is load-bearing for `sequences`: heavy and light use separate models, so mislabeling a chain produces a meaningless score. The threshold only affects the human / not-human classification, not the underlying score.

## Loose chains

- `sequences` (string[]) —
Antibody variable-region amino acid sequences, all of `chain_type`. One to 50 per job. Required unless you give `chain_pairs` — a request with neither is refused.

- `sequence_names` (string[]) —
Optional names for each sequence.

## Heavy/light pairs

- `chain_pairs` (object[]) —
Heavy/light pairs belonging to the same antibody. One to 50 per job. Each entry takes a required `heavy` sequence, a required `light` sequence (kappa or lambda), and an optional `name`.

Use this shape whenever you have both chains of one antibody, since the paired score is otherwise unavailable. `chain_type` does not apply here — each entry already says which sequence is which.

> **The pairing is your assertion, and nothing checks it**

> Humatch's paired classifier scores whatever combination it is handed. Nothing in the job can tell that two sequences came from the same antibody, so a mis-paired request returns a confident paired score for a chimera that does not exist. Pair the chains correctly before you submit; the output will not warn you.

## Chain type and scheme

- `chain_type` (string) —
Chain type of every entry in `sequences`: `H` for heavy, `L` for light (kappa or lambda). Selects the matching humanness model. It does not apply to `chain_pairs`.

- `scheme` (string) —
Numbering scheme requested for alignment: `imgt`, `chothia`, or `kabat`. ⚠ Humatch always aligns to IMGT, so a `chothia` or `kabat` request is **recorded and not applied** — the result reports it back in `requested_scheme` while `numbering_scheme` stays `imgt`. Position numbers in the output are IMGT whatever you send. For Chothia or Kabat numbering, use [ANARCI](/docs/anarci).

## Classification threshold

- `score_threshold` (float) —
Humanness score at or above which a sequence is classified as human (0.0–1.0).

## Read next