We are actively developing our finetuning pipelines. Please email info@tamarind.bio if any of the below use cases sound interesting to you, and we can work with you to determine the best approach given your available data.

​
Language model for property prediction

We offer protein language model finetuning using ESM2 for both regression and classification. Simply upload a CSV containing a column for your protein sequences and a column for your property of interest, and we’ll create a custom model. This model will then appear in your My Models tab, and you can use it for inference with your new protein sequences.

​
Antibody property prediction

For antibody datasets with a heavy and light chain, we offer an Antibody finetuning tool, originally trained on specificity data, which uses embeddings from Antiberty for regression and classification. Select the columns for your sequence, property, and job name, and we’ll create a custom model for your use case.

​
Point mutations

It’s difficult for current state of the art models to capture differences in point mutations. We recommend ALDE if you have some point mutations and want to find what others to test.

​
Structure Prediction

We offer a tool to finetune Boltz on your own custom cifs and use it for inference.

​
Large Binding Affinity Datasets (30,000+ data points)

We recommend AlphaBind if you have large data sets of binding affinity data - reach out to learn more.