An overview to the predict-phytolrr program

The predict-phytolrr program could identify LRRs effectively, especially for LRR motifs in plant LRR-RLKs.

The program is based on the position specific scoring matrix (PSSM) algorithm and its training dataset are 4000 LRR motifs highly conserved sequences (HCSs) from LRR-RLKs of 17 land plant species. Please find detailed information at https://www.phytolrr.com/about.

Install the program

predict-phytolrr can be installed using pip ONLY under Python3 environment:

$ pip install predict-phytolrr

A virtual environment(virtutalenv e.g.) is recommended to prevent the damage of the system-wide Python environment.

Run the program

You can run the program in two ways:

• run predict-phytolrr from the command line
• import the phytolrr_predictor module in the python source code

Predict Plant LRRs from the command line

The predict-phytolrr could accept sequence(s) in either directly input from the command line or in fasta file(s), predict the LRRs, and print out the results either in the command line or in html files.

Two ways to feed the sequences

First way: The sequence can be written directly through the command line:

$ predict-phytolrr MKLPHLLPFLTLILFAFAFSMTLPLMSESHLSLLNNRTDQQ....(very long)
Prediction result for seq Unknown_from_cmd_line:
LRR offset 8, FLTLILFAFAFSMTLP, score 5.68619738546084
LRR offset 81, VTYLNLTHTGLQGTLT, score 17.662632191299593
LRR offset 105, LRVLALRNNSLQGSIP, score 36.37469484306763
LRR offset 129, LQVLRVSQNQLEGKIP, score 35.079609050990754
LRR offset 153, LQRLILSYNRLEGSIP, score 39.13663517011341
... many other results followed

Second way: The sequences can also be read from fasta files. Assuming the file test.fasta contains one or more sequences, to predict LRRs for all the sequences in the file, use the -f option to specify a fasta file:

$ predict-phytolrr -f test.fasta
The sequence AL3G49070.t1 contains unexpected amino *, will not be predicted
The sequence AL3G41990.t1 contains unexpected amino *, will not be predicted
The sequence AL8G14250.t1 contains unexpected amino *, will not be predicted
Prediction result for seq AL5G26370.t1:
LRR offset 82, VTGVDLGGLKLTGVVS, score 20.609043937370146
LRR offset 106, LRSLNLADNFFRGAIP, score 33.523950081875626
LRR offset 130, LQYLNMSNNFLGGVIP, score 34.2935452040218
LRR offset 154, LSTLDLSSNHLEQGVP, score 29.383744295363456
LRR offset 178, LVILSLGRNNLTGKFP, score 34.567366699698404
... other predicted LRRs for seq AL5G26370.t1

Prediction result for seq AL5G26330.t1:
LRR offset 74, VTRLDLGGLQLGGVIS, score 24.76974498840441
LRR offset 98, LISLNLYDNSFGGTIP, score 34.4857875432728
LRR offset 122, LQHLNMSYNFLGGGIP, score 31.754313744552327
LRR offset 146, LLELDLISNHLGHCVP, score 10.059227438374975

... many other results followed

More than one fasta files could also be accepted by using the -f option for each fasta file:

$ predict-phytolrr -f test1.fasta -f test2.fasta

... all prediction results for all sequences in test1.fasta and test2.fasta

NOTE

1. Sequences with undefined amino acids read from fasta files, for instance "X", "*", will be skipped in the prediction process.
2. The asterisk "*" at the end of the fasta sequence will be removed automatically during the prediction process.

To display the results in html

The result can also be dumped to HTML files, so that they can be viewed in a browser. Using the -p option to specify the root path of the resulting HTML files:

$ predict-phytolrr -f test1.fasta -p ./
Warning: The sequence AL3G49070.t1 contains unexpected amino *, will not be predicted
Warning: The sequence AL3G41990.t1 contains unexpected amino *, will not be predicted
Warning: The sequence AL8G14250.t1 contains unexpected amino *, will not be predicted
214 sequences were predicted and the results have been saved in the file results.html

The program will generate two files:

• results.js to save the prediction results
• results.html to show the prediction results

To specify a self-defined training dataset(motifs)

If you have some motifs, and you want to predict LRRs based on the PSSM matrix which is generated by the motifs, you can use the option -t/--train-motifs-file to specify a path to a file which contains self-defined training-motifs. The training-motifs-file should contains multiple motifs, and each motif per line with the same length. The motifs length is not necessary to be 16. For example, the content of the training-motifs-file may like:

LEVLFLHGNQL
VTYLNLTHTGL
LQIFSIGGCHI
LKLLHLHHNFV
(More motifs follow)

The length of the predicted motifs will be of the same length with the motifs of the training dataset.

Predict Phyto LRRs in source code

The predict-phytolrr also provides an API interface, so that it can be easily integrated into other python projects.

To use predict-phtolrr to predict phyto-LRRs, just import phytolrr_predictor and call predict, which is defined as:

def predict(seq, matrix=None):
    """Predict motifs from the `seq` str. If the `matrix` is None, using the built-in matrix to predict motifs"""

Code example:

>>> import phytolrr_predictor
>>> test_seq = 'MGILFFLFALTLTLSSLSSSVFGLTQDGEALLEMKRGLNDTKGLLSNWKDTDINPCNWTRISCHLHDQRVRVINLPFLRLGGTISPSIGKITRLHRLAIH'
>>> ms = phytolrr_predictor.predict(test_seq)
>>> print(len(ms))
2
>>> print(ms[0].offset)                             # Print the offset of the first predicted motif 
6
>>> print(test_seq[ms[0].offset:ms[0].offset+16])   # Print the first predicted motif
LFALTLTLSSLSSSVF
>>> print(ms[0].score)
10.774822491308072
>>> print(ms[0].probability)
0.0005707622379645206
>>> print(ms[0].fdr_probability)
0.04882352941176471

To use self-defined matrix to predict motifs, the matrix should be generated using function pssm_matrix.calc_pssm_matrix:

def calc_pssm_matrix(motif_seqs_str):
    """Generate the PSSM matrix from baseline motifs `motif_seqs_str`"""

Code example:

>>> import phytolrr_predictor
>>> from phytolrr_predictor.tools import pssm_matrix
>>> baseline = [
'LEVLFLHGNQLENDPY',
'VTYLNLTHTGLQGTLT',
'LQIFSIGGCHIKGSIP', ....]
>>> matrix = pssm_matrix.calc_pssm_matrix(baseline)
>>> ms = phytolrr_predictor.predict(test_seq, matrix)
# other operations...