Pairing and combinations

Pairing determines which inputs belong to the same job. Preserve existing experiments with row mode, or deliberately generate new experiments from independent lists.

One row per job

This is the default for CSV batches. Inputs and settings stay attached to their original row. Empty optional inputs remain absent.

For example, rows containing H1/L1 and H2 with no light chain create exactly those two jobs. Use this mode for antibody pairs, complete folding complexes and file manifests with different settings.

Pairwise lists

Pairwise lists (1:1) treats each numbered CSV input column as an independent ordered list. The first values run together, then the second values, and so on.

• Equal lists: A1, A2 paired with B1, B2 creates A1/B1 and A2/B2.
• Single shared value: A1 paired with B1, B2 creates A1/B1 and A1/B2.
• Unequal lists: Two values paired with three values is an error. Every nonempty list must contain either one value or the longest list's count.

Empty CSV cells are removed from lists. They do not preserve a missing partner's position. Keep One row per job when blank cells represent optional members of existing experiments.

All combinations

All combinations of input columns creates the Cartesian product. Two proteins and three ligands create six jobs. Each populated numbered input column is an independent factor, including separate protein columns.

This does not swap input roles or automatically remove self-comparisons. For selected pairs, symmetric comparisons without duplicates, or exclusions, write the intended jobs explicitly in row mode.

Review generated jobs

List modes generate job names. Supported attributes, such as copies or an attached MSA, travel with their input. Setting columns require row mode so experimental settings are not detached from their intended inputs.

Duplicate inputs remain separate entries. Inspect Preview and the credit estimate, especially before choosing all combinations. Expansion above 250 jobs is rejected.

Boltz-2 molecule fields

Each field supplies one member of the complex; its lines or FASTA records are alternatives. Protein 1 and Protein 2 stay together in each prediction. With two alternatives in each field, Batch pairing → Pairwise creates two jobs; All combinations creates four. A field with one entry is shared across jobs. Templates and standalone MSAs are shared too.

Use FASTA headers when one sequence wraps across multiple lines. Without headers, each nonempty pasted sequence line is a separate alternative. Other tools retain their own collection rules.

To make all records in a FASTA input members of the same complex, choose FASTA records → Members of one complex beside that field. Each record becomes a numbered field. Existing fields remain in the experiment, and their alternatives still follow your pairing choice.