Input Parameters

Either retrieve the cDNA sequence using the ensembl transcript ID or directly paste your sequence in the respective text area.

Foundation Models

Evolutionary scale modeling (ESM2) is a protein language model designed based on the bidirectional encoder representations from transformers (BERT) framework and was pretrained on the Uniprot dataset using masked language model learning. Nucleotide Transformers (NT) are state-of-the-art foundation models pretrained on DNA sequences from multiple species. These models yield context-specific representations of nucleotide sequences, which allow for accurate predictions on multiple genomic applications. In this study, the ESM2 and NT models were fine tuned to predict both cognate and near cognate TIS. A meta model was also created based on the final outputs from both models.

Results

A csv file will be returned containing the following columns:

Column Name	Description
Start Codon	Start Codon used in candidate TIS.
Reading Frame	Reading Frame where the start site is located.
Start Site	Index of start site with respect to the transcript.
NT	Nucleotide Transformer Output.
ESM2	ESM2 output.
MetaTIS	Meta model output.