LICHEN

Light-chain Immunoglobulin sequence generation Conditioned on the Heavy chain and Experimental Needs

LICHEN is a Seq2Seq language model designed to predict a light chain given a heavy.
Use the form below to input your heavy sequence and customise the prediction settings.
The model has been trained on paired OAS human datasets and therefore its predictions will reflect the underlying biases in this data, hence we provide users the options to customise the output chains.

The LICHEN Python package is freely available on Github.

For a full description of the pipeline, or if you use this software, please refer to: Capel et al. 2025

NB: We do not allow users to process more than 5 sequences per file on the webtool, if you are processing large numbers of sequences please download the latest version and run locally. GitHub

Heavy chain sequence:

load example

How many light chains do you want to generate?

Please choose either:

You may input custom CDR Sequences, or else leave blank:

Definitions:

IMGT Kabat

CDRL1

CDRL2

CDRL3

Filter generated sequences to ensure they pass any, or all of, the following checks:

We have temporarily disabled these filters, but they will be re-enabled in the future as we move our webapps to a GPU server. If you require these filters, please run the LICHEN package locally Github.

Redundancy Diversity AbLang2 ANARCII Humatch