CasTER (Cas9 Thermodynamic Energy Repair prediction) is a tool to predict the outcome of cell-internal repair machinery after Cas9-induced DNA double strand breaks. It uses a tree-based model based on thermodynamic features generated using MOSAICS. The automatic generation of PDB files from DNA sequences is done using fd_helix.c.
CasTER is developed by Florian Störtz at the group of Dr. Minary, University of Oxford.
Running CasTER requires two steps:
python load_data.py
Positional arguments are:
- oligo file: path to input file of tab separated oligos, containing the following columns: ID, Guide Sequence, Target Sequence, Scaffold, Subset, Comments, PAM Index, Strand. Only ID, Target Sequence, PAM Index and Strand need to have meaningful content. For an example, see
oligos.txt - number of CPU workers (default: 48 - choose this proportional to the number of CPUs available)
- width in base pairs around the cut site which is rendered and used by MOSAICS (default: 21)
- indel file: path to a file of measured indels, each oligo introduced by
@@@ID, then following an enumeration of indels with tab-separated columns: indel signature, read count, repaired sequence. For an example, seeindels.txt. This file does not need to be supplied, but if it is supplied then CasTER calculates benchmarks (KL divergence, Pearson r, Spearman r) based on the difference between true and predicted indel distributions.
This generates the mosaics_data.pickle file containing the MOSAICS energies which are needed for prediction in the following step.
To run a prediction, call
python caster.py
Positional arguments are:
- sum_level:
0: predict indel length,1: predict indel position,None: predict indel length and position - feature_set:
0: use sequence and energy features,1: use energy features only,2: use sequence features only - mosaics_data_path: path to the file containing the MOSAICS energies (default:
mosaics_data.pickle)
Running python caster.py gives you an example output using precalculated MOSAICS energy features.
data_FORECasT.xz contains precalculated MOSAICS energy features for 7000 oligos within the FORECasT data set, which CasTER uses as a basis. Therefore, CasTER can be run out-of-the-box on FORECasT data.
Download MOSAICS and install using the instructions given there. Move mosaics.x to the CasTER directory.
Adjust the paths to the MOSAICS library folder in input
See requirements.txt