Input
RBS Predictor
Enter 5'-UTR sequences
(17 nt upstream of start codon)
or
Upload a file
(*.csv, *.txt)
The RBS Predictor is a neural network ensemble that predicts the rTR of RBSs based on 5’-UTR sequences as queries, which can be added in the sequence window or via file input above.

User manual

The RBS Predictor based on a deep-learning model (SAPIENS) that was trained on approximately 250,000 experimentally characterized RBS sequences. It accepts 5’-UTR sequences as an input and predicts their relative translation rate (rTR) with high confidence (R2 = 93%).
Input
The RBS Predictor DNA or RNA sequence inputs following the (IUPAC base code), which can be specified line by line via the sequence window or file upload above. The 5’-UTR sequence directly before the start codon of the CDS, whose expression is to be predicted, must be specified. To obtain most accurate predictions, we recommended to add at least 17 nt, but also shorter queries (min. 10 nt) are accepted (e.g. if the transcription start lies closer to the start codon). Note that sequences shorter than 17 nt might lead to lower prediction accuracy.
Example regular query:
ACCACAGAGTTGAGAGG[ATG]
ACGTCGAGGAGGATATT[ATG]
ACCCCTGTCTGTGCTGA[ATG]
[three sequences, 17 nt; do not add the [ATG] start codon]
Alternative to individual sequences (containing A, C, G and T/U), degenerate sequence positions coding more than one base can be specified using the (IUPAC code). In this case, RBS Predictor will convert the degenerate query into all encoded non-degenerate sequences performing predictions for all of them. This allows to perform up to 65,536 (48 for N8) predictions per job in batch.
Example degenerate query:
ACCACAGAGNRGAGAGG ACCACAGAGAAGAGAGG
ACCACAGAGAGGAGAGG
ACCACAGAGCAGAGAGG
ACCACAGAGCGGAGAGG
ACCACAGAGGAGAGAGG
ACCACAGAGGGGAGAGG
ACCACAGAGTAGAGAGG
ACCACAGAGTTGAGAGG
Output
As main output, RBS Predictor returns the (rTR) of each query sequence, which represents the relative RBS strength on a scale from 0 to 100,000 arbitrary units (a.u.). The value corresponds to the cell-specific accumulation rate of an sfGFP-labelled reporter protein during a period of 290 minutes after transcriptional induction. As secondary output, RBS Predictor specifies the (uncertainty) of each prediction in percent of the full rTR-scale (i.e. 1% uncertainty corresponds to ±1,000 a.u.). This parameter can additionally help users to select RBS sequences, for which the model is “most sure” about its predictions (< 3% uncertainty are recommended).
Example output (*.txt):
# sequence rTR_(a.u.) uncertainty_(%)
ACCACAGAGTTGAGAGG 1431.706 0.419
ACGTCGAGGAGGATATT 95347.457 0.204
ACCCCTGTCTGTGCTGA 46.541 0.015
[in this example, an intermediate, a very strong, and a weak RBS with prediction uncertainties well below 1% are displayed]

When you use the RBS Predictor in your published work, please do not forget to cite:

  • Höllerer, S., Papaxanthos, L., Gumpinger, A. C., Fischer, K., Beisel, C., Borgwardt, K., Benenson, Y., & Jeschek, M. (2020). Large-scale DNA-based phenotypic recording and deep learning enable highly accurate sequence-function mapping. Nature communications, 11(1), 3551 (https://doi.org/10.1038/s41467-020-17222-4)

Usage data
Queue length
0
Summary statistics
Jobs this month
3
Jobs all time
195
Registered users
8