Instructions for TCRdenoise-1.0
Input format
- The server only accepts amino acid sequences takes in newline separated TCR sequences. The input file should contain the 6 CDR loops and their target peptide, comma-separated. The input file can also contain a validation column.
- Note that the header names need to be: "A1,A2,A3,B1,B2,B3,peptide,validation", extra columns in your dataset will be ignored and will not cause errors to the results.
Submission
- Paste the CDR and peptide sequence(s) into the box
- Load an test example input or upload a file from your local machine. It is possible to either upload a file or load an example. Please refer to the input format description;
- Upload a training set from your local machine. Also this file should follow the input format
- If figures of the peptide-specific silhouette score, clustering solution and venn diagrams are wanted tick "Include Figures".
- If matrices generated by TCRdist3 and TCRbase are needed tick "Include Matrix".
Click the submit button when the data has been inputted.
Output
After the server successfully finishes the job, a Server Output page shows up.
Computational time can range from a couple of seconds to several minutes depending on the queue, the sample size, and wether Matrix and figures are included.
The output contains a summary of the amount of TCRs classified as binder or noise both in total and per peptide, and link to downloading the files generated: consensus, matrix, silhouette curves, clustering solutions, and venn diagrams.
Output of TCRdenoise-1.0
After submitting a data set to TCRdenoise-1.0, an output will be shown, depending on the setting chosen.
- A summary is shown of the amount of binders and noise classified by TCRdenoise-1.0 both per peptides and in total.
- "Consensus", a file containing the input data along with the classification results from each method and their consensus. 0 indicates noise (non-binder), 1 indicates a binder for each TCR.
- "TCRdist": Classification based on TCRdist3.
- "TCRbase": Classificaiton based on TCRbase.
- "consensus_denoise": classification based on the consensus of TCRdist3 and TCRbase.
- If "Included Matrix" was selected:"Matrix", a .zip file containing each peptides distance matrix generated by TCRdist3 and TCRbase.
- If "Included Figures" was selected: multiple .zip files containing figures of: silhouette curves, clustering solution and venn diagram for each peptide.
The output can be downloaded as either a .csv file or a .zip file.
Description
TCRdenoise-1.0 is a sequence-based method for separating binding and non-binding TCRs in a peptide-specific repertoire. It uses TCRbase [1] and TCRdist3 [2] to compute pairwise TCR distance matrices, from which clustering solutions are generated using agglomerative clustering across a range of distance thresholds. The clustering solutions are then evaluated using a refined silhouette score modified to handle singletons to identify the optimal distance threshold for each distance matrix. Next, TCRs located in clusters are classified as binders, and non-clustered TCRs as noise (non-binders). Finally, the selected clustering solutions obtained from the two similarity metrics are combined into a consensus clustering.
[1]Montemurro, A., Jessen, L. E. & Nielsen, M. NetTCR-2.1: Lessons and guidance on how to develop models for TCR specificity predictions. Front. Immunol. 13, (2022).
[2]Mayer-Blackwell, K. et al. TCR meta-clonotypes for biomarker discovery with tcrdist3 enabled identification of public, HLA-restricted clusters of SARS-CoV-2 TCRs. eLife 10, e68605 (2021)
