Input Data

OpenKinetics Predictor accepts reaction data as CSV in the web interface. The public API uses the same column names when data is submitted as JSON.

Column names are case-sensitive. Empty CSV rows are ignored.

Required columns

Every input row must contain a Protein Sequence value and either a Substrate value or a Substrates value.

• Protein Sequence: full amino-acid sequence. Use uppercase one-letter amino-acid codes from ACDEFGHIKLMNPQRSTVWY.

• Substrate: one SMILES or InChI string.

• Substrates: one or more SMILES or InChI strings separated by semicolons.

• Products: one or more product SMILES or InChI strings separated by semicolons. This column is only valid with Substrates.

Do not include both Substrate and Substrates in the same CSV. If a Products column is present, the frontend treats the file as a full-reaction CSV.

CSV formats

Single-substrate CSV

Use this format for one protein/substrate pair per row.

Required columns:

• Protein Sequence

• Substrate

Example:

Protein Sequence,Substrate
MKTLLILAV...,CC(=O)O

This is the native format for CataPro, DLKcat, EITLEM-Kinetics, IECata, KinForm-H, KinForm-L, MMISA-KM, OmniESI, OmniESI + O2DENet, RealKcat, UniKP, and CatPred Km.

Multi-substrate CSV

Use this format when one reaction row contains an ordered list of substrates.

Required columns:

• Protein Sequence

• Substrates

Example:

Protein Sequence,Substrates
MKTLLILAV...,CC(=O)O;O;C1CCCCC1

CatPred kcat consumes Substrates natively as one combined substrate set. Single-substrate methods can also run on this format by expanding each semicolon-separated substrate independently.

Full-reaction CSV

Use this format for methods that need both substrates and products.

Required columns:

• Protein Sequence

• Substrates

• Products

Example:

Protein Sequence,Substrates,Products
MKTLLILAV...,CC(=O)O;O,C(C(=O)O)O

TurNup kcat requires this format. Other compatible methods can accept a full-reaction file, but only TurNup uses the Products values for prediction. Supplied products are still validated.

Method compatibility

The frontend builds the available method list from /api/v1/methods/. The table below reflects those method descriptors.

For methods whose minimum format is Protein Sequence + Substrate, the frontend also accepts Substrates and full-reaction files by adapting each listed substrate to the method’s single-substrate contract.

Method	Targets	Minimum or native input columns	Max sequence length
CataPro	kcat, Km, kcat/Km	Protein Sequence, Substrate	1,000
CatPred	kcat, Km	kcat: Protein Sequence, Substrates; Km: Protein Sequence, Substrate or Substrates	2,048
DLKcat	kcat	Protein Sequence, Substrate	No method limit
EITLEM-Kinetics	kcat, Km	Protein Sequence, Substrate	1,024
IECata	kcat/Km	Protein Sequence, Substrate	1,000
KinForm-H	kcat, Km	Protein Sequence, Substrate	1,500
KinForm-L	kcat	Protein Sequence, Substrate	1,500
MMISA-KM	Km	Protein Sequence, Substrate	500
OmniESI	kcat, Km	Protein Sequence, Substrate	1,000
OmniESI + O2DENet	kcat, Km	Protein Sequence, Substrate	1,000
RealKcat	kcat, Km	Protein Sequence, Substrate	1,022
TurNup	kcat	Protein Sequence, Substrates, Products	1,024
UniKP	kcat, Km	Protein Sequence, Substrate	1,000

Semicolon lists

Substrates and Products use semicolons to separate molecules. Empty fragments are ignored, so A;;B is read as A and B.

For single-substrate methods running on a Substrates list:

• kcat output is reduced to the maximum successful per-substrate prediction.

• Km and kcat/Km outputs remain ordered JSON arrays, one value per substrate.

• Failed per-substrate positions are recorded as null in the array and in the result Extra Info.

Protein Sequence may also contain semicolon-separated candidate sequences. The output remains one row per input row:

• kcat uses the maximum successful candidate-sequence prediction.

• Km uses the minimum successful candidate-sequence prediction.

• Direct kcat/Km uses the maximum successful candidate-sequence prediction.

• When both candidate sequences and Substrates are present, Km and kcat/Km keep ordered per-substrate arrays.

Validation

The frontend checks CSV structure as soon as a file is selected. A valid file is labelled as single-substrate, multi-substrate, or full-reaction and the compatible method selectors are enabled.

The optional validation step checks:

• missing Protein Sequence values;

• invalid amino-acid characters;

• invalid substrate or product SMILES/InChI strings;

• empty substrate or product lists;

• per-method protein sequence length limits;

• the overall server sequence limit of 10,000 residues;

• optional sequence similarity to method training data.

Invalid substrates or proteins are skipped during prediction. Invalid Products values block submission because full-reaction methods need valid product chemistry.

Long sequences

When validation finds sequence-length warnings, the frontend lets you choose how submission should handle long sequences:

• truncate: default. The service keeps the first and last portions of the sequence. For example, with a 1,000-residue limit, a 1,200-residue sequence becomes the first 500 residues plus the last 500 residues.

• skip: exclude rows or sequence candidates that exceed the applicable limit.

For multi-target jobs, the applied limit is the strictest selected method limit, capped by the 10,000-residue server limit.

Submission options

The method-selection footer exposes input-related switches:

• Canonicalize substrates: enabled by default. Equivalent SMILES are normalized before prediction. Disable it to use each method’s native preprocessing.

• Prefer experimental data: when a curated match from BRENDA, SABIO-RK, or UniProt is found, return that value instead of a model prediction.

• Include similarity columns: enabled by default. Adds kcat training-set similarity columns to the output CSV.

Templates

The frontend provides downloadable CSV templates:

• /templates/single_substrate_template.csv

• /templates/multi_substrate_template.csv

• /templates/full_reaction_template.csv