piezo

Predict the effect of a genetic mutation on the effect of an antibiotic using a supplied AMR catalogue.

This code was developed as part of the CRyPTIC international tuberculosis consortium. If you would like to use the software commercially, please consult the LICENCE file.

Pre-requisites

Everything is Python3. Everything else is fairly standard / should be installed by the setup.py if you don't already have it

• pandas
• ujson
• pytest
• pytest-cov

Installation

First clone/download the repo. Then

$ cd piezo
$ ls
$ python setup.py develop --user

Included files

$ $ ls tests/test-catalogue/
NC_004148.2.gbk                    NC_004148.2_TEST_GM1_RFUS_v1.0.csv

NC_004148 is the reference genome of the human metapneumovirus and is used primarily for unit testing since it is small and fast to parse.

Design of AMR catalogue

piezo is written so as to be extendable in the future to other ways of describing genetic variation with respect to a reference. It includes the concept of a grammar which specifies how the genetic variation is described.

At present only a single grammar, GARC1 is supported. GARC is short for Grammar for Antimicrobial Resistance Catalogues. This grammar is described in more detail elsewhere, however in brief, it is a gene-centric view (and therefore has no way of describing genetic variation that lies outside a coding region, other than as a 'promoter' mutation). All mutations start with the gene (or locus) name which must match the name of a gene (or locus) in the relevant GenBank file. It is the user's responsibility to ensure this, although e.g. the gumpy package can be used to perform such sanity checks. The mutation is delineated from the gene using a @ symbol and within the mutation _ is used as a field separator to separate the different components. All variation is described as either a SNP or an INDEL. If they occur within a coding region SNPs are specified by their effect on the amino acids which are always in UPPERCASE e.g. rpoB@S450L. If in the assumed promoter region, then the nucleotide change and position is specified e.g. fabG1@c-15t. Nucleotides are always in lowercase. INDELs can be specified at different levels of granularity e.g. rpoB@1250_indel means 'any insertion of deletion at this position', but we could equally be highly specific and say rpoB@1250_ins_cta which means 'an insertion of cta at this position'. There is also the special case of frameshifting mutations which are described by fs.

Wildcards are also supported. Hence rpoB@*? means 'any non-synoymous mutation in the coding region of the protein'. To avoid confusion the stop codon is represented by ! which is non-standard. Het calls are, at present, represented by a Z or z depending on whether they occur in the coding or promoter regions. This may be extended in the future. Likewise null calls are represented by an X or x.

The general principle is each mutation can 'hit' multiple rules in the catalogue, but it is the most specific rule that will be followed. Hence consider a toy example, again from TB

rpoB@*?     RIF   U   any non-synoymous mutation in the coding region has an unknown effect of RIF
rpoB@S450?  RIF   R   any non-synoymous mutation at Ser450 confers resistance
rpoB@S450Z  RIF   F   a het call at Ser450 should be reported as an F (fail).

Example

A demonstration script called piezo-predict.py can be found in the bin/ folder of the repository which following installation should be in your $PATH. A made-up catalogue for testing purposes can be found in tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv which is based on the Human metapneumovirus, however the entries are fictious. It contains two drugs and a series of mutations in the M2 gene.

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@L73L
{'DRUG_B': 'S', 'DRUG_A': 'S'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@L73R
{'DRUG_A': 'R', 'DRUG_B': 'U'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@L73Z
{'DRUG_B': 'S', 'DRUG_A': 'F'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@300_indel
{'DRUG_B': 'U', 'DRUG_A': 'U'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@300_ins
{'DRUG_B': 'U', 'DRUG_A': 'U'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@300_ins_2
{'DRUG_B': 'U', 'DRUG_A': 'U'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@300_ins_3
{'DRUG_A': 'U', 'DRUG_B': 'R'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@300_ins_4
{'DRUG_B': 'U', 'DRUG_A': 'U'}

$ piezo-predict.py --catalogue tests/test-catalogue/NC_004148.2_TEST_v1.0_GARC1_RFUS.csv --mutation M2@300_ins_cta
{'DRUG_B': 'R', 'DRUG_A': 'U'}