PKSpop

A pipeline to predict type I PKSs protein order in polyketide biosynthetic assembly lines.

PKSpop comprises three main steps to infer protein order:

1. Identify class memberships for query docking domains and align the sequences
2. Pair each class I Ndd with all class I Cdds and use Ouroboros to predict the interaction probability for each pair. The probabilities are filled into a matrix
3. Infer protein order from the probability matrix by the Hungarian algorithm, a global optimization method, which finds a match between Ndds and Cdds that maximizes the overall interaction probability. The inference method takes the assembly line constraints and compatibility class into account.

Run PKSpop

python PKSpop/code/run_analysis.py input_json_file.json

Input

The input of PKSpop is a JSON file which contains following informaion:

• gbk_path: path where the antiSMASH .gbk file of the query PKS gene cluster
• protein_id: list of identifiers of query proteins whose order will be predicted
• id_category: the category of the protein identifiers: "gene", "protein_id" or "locus_tag"
• result_path: path where result will be saved
• Ouroboros_path: path to Ouroboros repositry
• Ouroboros_int_frac: list of int_frac parameter of Ouroboros with default [0.9, 0.8]. It is recommended to add numbers below 0.8 if there are more than 10 query proteins.
• n_repeat: number of repeat time to run Ouroboros with each int_frac parameter An example can be found in data/test

Output

The prediction results are in result_path/output/

• protein_order_prediction.txt gives the predicted order of the query assembly line
• protein_interaction_prediction.txt gives the predicted interacting protein pairs in the query assembly line
• int_prob_mtx.csv is the pairwise interaction probabilities matrix of query proteins predicted by Ouroboros

Additional files used in prediction process:

• dd_raw.fasta is the raw sequences extracted from the input .gbk file
• dd_class_*.fasta is the sequences of 3 compatibility classes
• dd_hmmscan_oupt.txt is the result of hmmscan, which contains the class information of the sequences
• dd_class_1_aln.afa is the aligned class 1 sequences
• dd_class_1.afa is the conserved region on the aligned sequences
• dd_class_1_paired.afa is the paired sequences of all query proteins
• dd_class_1_ouro_inpt.fasta is the fasta file that input into Ouroboros
• Ouroboros_class_1_ouro_inpt_soft_warm is the Ouroboros' output

License

This project is licensed under the BSD-3 license. See the LICENSE file for details.

Requirements

PKSpop requires Python 3.6+. The following tools should be installed/downloaded before running PKSpop:

• Ouroboros
• HMMER

Packages:

• Biopython
• NumPy
• Pandas
• SciPy