Manual of beRBP webserver

beRBP is an approach to predict RBP-RNA interaction given RNA sequence(s) and PWM(s) of RBP(s).



beRBP requires two types of input: RNA sequence(s) and RBP:PWM(s) model. Email address is optional. If an email is provided, a notification will be sent to the email once the job is finished.

1. RNA sequence(s)

2. RBP:PWM(s) Model

beRBP provides three types of RBP:PWM(s) models: Specific, General (prebuilt-in PWMs or user-provided PWMs), or PWMs inference from a RBP sequence based on RNA Binding Domain (RBD) identity.

A. Specific model

There are 37 beRBP-Specific models for 29 RBPs. Each model is trained by the known targets of the RBP itself. For example, CIRBP model is trained by the known targets of CIRBP.


B. General model

beRBP-General provides a universal classifier for all RBPs with known PWMs, which is trained by pooling all known RBP-RNA interactions. Users can select the prebuilt-in 143 RBPs/175 PWMs or upload your own PWM(s).

B.1 Select Prebuilt-in RBP:PWM(s)

Step 1 of 2:


Step 2 of 2:


B.2 Upload your own PWM


C. PWM(s) inference from a RBP sequence

If your RBP of interest doesn’t have a PWM, beRBP can predict PWM(s) based on the findings that two RBPs for which their RBDs (RNA binding domain) are >50% identical are likely to have similar motifs (Ray, Kazan et al. 2013). Pfam models used to scan for RBDs are obtained from supplementary data 9 in Ray’s paper (Ray, Kazan et al. 2013). Based on the predicted PWM(s), beRBP will use the General model to predict RBP-RNA interactions.

Step 1 of 2: Provide one RBP sequence to predict PWM(s)


Step 2 of 2: Select inferred PWM(s) to continue RBP-RNA interaction prediction





Compared to DeepBind and RBPmap, beRBP expands the application scope by providing predictions for 143 RBPs and enabling predictions for user-uploaded PWMs or RBP sequences.
#: RBPmap requires the PWM length to be within 4~10bp, while beRBP has no restriction on the motif length.
*: Among the census of human RBPs (Gerstberger et al., 2014, Nat Rev Genet), PWMs of 169 RBPs (click for sequences) can be predicted.


Approximate runtime:
*The runtime was estimated on a single job with one RBP-PWM model and one RNA sequence (~1050nt) except the prediction from one RBP sequence, where 4 PWMs detected from the RBP sequence were used.