def test_aniblastall_concordance(): """Test concordance of ANIblastall method with JSpecies output.""" # Make/check output directory mode = "ANIblastall" outdirname = delete_and_remake_outdir(mode) # Get dataframes of JSpecies output aniblastall_jspecies = parse_table(JSPECIES_OUTFILE, 'ANIb') # Identify our input files, and the total lengths of each organism seq infiles = pyani_files.get_fasta_files(INDIRNAME) org_lengths = pyani_files.get_sequence_lengths(infiles) # Test ANIblastall concordance: # Make fragments fragfiles, fraglengths = anib.fragment_fasta_files(infiles, outdirname, pyani_config.FRAGSIZE) # Build jobgraph jobgraph = anib.make_job_graph( infiles, fragfiles, anib.make_blastcmd_builder("ANIblastall", outdirname)) print("\nJobgraph:\n", jobgraph) print("\nJob 0:\n", jobgraph[0].script) # Run jobgraph with multiprocessing run_dependency_graph(jobgraph) print("Ran multiprocessing jobs") # Process BLAST; the pid data is in anib_data[1] aniblastall_data = anib.process_blast(outdirname, org_lengths, fraglengths, mode="ANIblastall") aniblastall_pid = \ aniblastall_data.percentage_identity.sort_index(axis=0).\ sort_index(axis=1) * 100. index, columns = aniblastall_pid.index, aniblastall_pid.columns diffmat = aniblastall_pid.as_matrix() - aniblastall_jspecies.as_matrix() aniblastall_diff = pd.DataFrame(diffmat, index=index, columns=columns) # Write dataframes to file, for reference aniblastall_pid.to_csv(os.path.join(outdirname, 'ANIblastall_pid.tab'), sep='\t') aniblastall_jspecies.to_csv(os.path.join(outdirname, 'ANIblastall_jspecies.tab'), sep='\t') aniblastall_diff.to_csv(os.path.join(outdirname, 'ANIblastall_diff.tab'), sep='\t') print("ANIblastall concordance test output placed in %s" % outdirname) print("ANIblastall PID:\n", aniblastall_pid) print("ANIblastall JSpecies:\n", aniblastall_jspecies) print("ANIblastall diff:\n", aniblastall_diff) # We'd like the absolute difference reported to be < ANIBLASTALL_THRESHOLD max_diff = aniblastall_diff.abs().values.max() print("Maximum difference for ANIblastall: %e" % max_diff) assert_less(max_diff, ANIB_THRESHOLD)
def run_anim_jobs(joblist: List[ComparisonJob], args: Namespace) -> None: """Pass ANIm nucmer jobs to the scheduler. :param joblist: list of ComparisonJob namedtuples :param args: command-line arguments for the run """ logger = logging.getLogger(__name__) if args.scheduler == "multiprocessing": logger.info("Running jobs with multiprocessing") if not args.workers: logger.debug("(using maximum number of worker threads)") else: logger.debug("(using %d worker threads, if available)", args.workers) cumval = run_mp.run_dependency_graph( [_.job for _ in joblist], workers=args.workers ) if cumval > 0: logger.error( "At least one NUCmer comparison failed. Please investigate (exiting)" ) raise PyaniException("Multiprocessing run failed in ANIm") logger.info("Multiprocessing run completed without error") else: logger.info("Running jobs with SGE") logger.debug("Setting jobarray group size to %d", args.sgegroupsize) run_sge.run_dependency_graph( [_.job for _ in joblist], jgprefix=args.jobprefix, sgegroupsize=args.sgegroupsize, sgeargs=args.sgeargs, )
def test_aniblastall_concordance( paths_concordance_fna, path_concordance_jspecies, tolerance_anib_hi, fragment_length, tmp_path, ): """Check ANIblastall results are concordant with JSpecies.""" # Get lengths of input genomes orglengths = pyani_files.get_sequence_lengths(paths_concordance_fna) # Perform ANIblastall on the input directory contents fragfiles, fraglengths = anib.fragment_fasta_files( paths_concordance_fna, tmp_path, fragment_length ) jobgraph = anib.make_job_graph( paths_concordance_fna, fragfiles, anib.make_blastcmd_builder("ANIblastall", tmp_path), ) assert 0 == run_mp.run_dependency_graph(jobgraph) # Jobs must run correctly # Process BLAST output result_pid = anib.process_blast( tmp_path, orglengths, fraglengths, mode="ANIblastall" ).percentage_identity # Compare JSpecies output to results result_pid = (result_pid.sort_index(axis=0).sort_index(axis=1) * 100.0).values tgt_pid = parse_jspecies(path_concordance_jspecies)["ANIb"].values assert result_pid - tgt_pid == pytest.approx(0, abs=tolerance_anib_hi)
def test_aniblastall_concordance(self): """ANIblastall results concordant with JSpecies.""" # Perform ANIblastall on the input directory contents outdir = os.path.join(self.outdir, "blastall") os.makedirs(outdir, exist_ok=True) fragfiles, fraglengths = anib.fragment_fasta_files( self.infiles, outdir, self.fragsize ) jobgraph = anib.make_job_graph( self.infiles, fragfiles, anib.make_blastcmd_builder("ANIblastall", outdir) ) assert_equal(0, run_mp.run_dependency_graph(jobgraph)) results = anib.process_blast( outdir, self.orglengths, fraglengths, mode="ANIblastall" ) result_pid = results.percentage_identity result_pid.to_csv(os.path.join(self.outdir, "pyani_aniblastall.tab"), sep="\t") # Compare JSpecies output to results result_pid = result_pid.sort_index(axis=0).sort_index(axis=1) * 100.0 diffmat = result_pid.values - self.target["ANIb"].values aniblastall_diff = pd.DataFrame( diffmat, index=result_pid.index, columns=result_pid.columns ) aniblastall_diff.to_csv( os.path.join(self.outdir, "pyani_aniblastall_diff.tab"), sep="\t" ) assert_less(aniblastall_diff.abs().values.max(), self.tolerance["ANIblastall"])
def test_dependency_graph_run(self): """Test that module runs dependency graph.""" fragresult = anib.fragment_fasta_files(self.infiles, self.outdir, self.fraglen) blastcmds = anib.make_blastcmd_builder("ANIb", self.outdir) jobgraph = anib.make_job_graph(self.infiles, fragresult[0], blastcmds) result = run_multiprocessing.run_dependency_graph(jobgraph) self.assertEqual(0, result)
def test_dependency_graph_run(self): """module runs dependency graph.""" fragresult = anib.fragment_fasta_files(self.infiles, self.outdir, self.fraglen) blastcmds = anib.make_blastcmd_builder("ANIb", self.outdir) jobgraph = anib.make_job_graph(self.infiles, fragresult[0], blastcmds) result = run_multiprocessing.run_dependency_graph(jobgraph) assert_equal(0, result)
def test_dependency_graph_run(path_fna_two, fragment_length, tmp_path): """Test that module runs dependency graph.""" fragresult = fragment_fasta_files(path_fna_two, tmp_path, fragment_length) blastcmds = make_blastcmd_builder("ANIb", tmp_path) jobgraph = make_job_graph(path_fna_two, fragresult[0], blastcmds) result = run_dependency_graph(jobgraph) assert 0 == result
def calculate_anim(infiles, org_lengths): """Returns ANIm result dataframes for files in input directory. - infiles - paths to each input file - org_lengths - dictionary of input sequence lengths, keyed by sequence Finds ANI by the ANIm method, as described in Richter et al (2009) Proc Natl Acad Sci USA 106: 19126-19131 doi:10.1073/pnas.0906412106. All FASTA format files (selected by suffix) in the input directory are compared against each other, pairwise, using NUCmer (which must be in the path). NUCmer output is stored in the output directory. The NUCmer .delta file output is parsed to obtain an alignment length and similarity error count for every unique region alignment between the two organisms, as represented by the sequences in the FASTA files. These are processed to give matrices of aligned sequence lengths, average nucleotide identity (ANI) percentages, coverage (aligned percentage of whole genome), and similarity error cound for each pairwise comparison. """ logger.info("Running ANIm") logger.info("Generating NUCmer command-lines") # Schedule NUCmer runs if not args.skip_nucmer: joblist = anim.generate_nucmer_jobs( infiles, args.outdirname, nucmer_exe=args.nucmer_exe, maxmatch=args.maxmatch ) if args.scheduler == "multiprocessing": logger.info("Running jobs with multiprocessing") cumval = run_mp.run_dependency_graph(joblist, verbose=args.verbose, logger=logger) logger.info("Cumulative return value: %d" % cumval) if 0 < cumval: logger.warning("At least one NUCmer comparison failed. " + "ANIm may fail.") else: logger.info("All multiprocessing jobs complete.") else: logger.info("Running jobs with SGE") run_sge.run_dependency_graph(joblist, verbose=args.verbose, logger=logger) else: logger.warning("Skipping NUCmer run (as instructed)!") # Process resulting .delta files logger.info("Processing NUCmer .delta files.") try: data = anim.process_deltadir(args.outdirname, org_lengths) except ZeroDivisionError: logger.error("One or more NUCmer output files has a problem.") if not args.skip_nucmer: if 0 < cumval: logger.error("This is possibly due to NUCmer run failure, " + "please investigate") else: logger.error( "This is possibly due to a NUCmer comparison " + "being too distant for use. Please consider " + "using the --maxmatch option." ) logger.error(last_exception()) return data
def test_aniblastall_concordance(self): """Check ANIblastall results are concordant with JSpecies.""" # Perform ANIblastall on the input directory contents outdir = self.outdir / "blastall" outdir.mkdir(exist_ok=True) fragfiles, fraglengths = anib.fragment_fasta_files( self.infiles, outdir, self.fragsize) jobgraph = anib.make_job_graph( self.infiles, fragfiles, anib.make_blastcmd_builder("ANIblastall", outdir)) self.assertEqual(0, run_mp.run_dependency_graph(jobgraph)) results = anib.process_blast(outdir, self.orglengths, fraglengths, mode="ANIblastall") result_pid = results.percentage_identity result_pid.to_csv(self.outdir / "pyani_aniblastall.tab", sep="\t") # Compare JSpecies output to results result_pid = result_pid.sort_index(axis=0).sort_index(axis=1) * 100.0 diffmat = result_pid.values - self.target["ANIb"].values aniblastall_diff = pd.DataFrame(diffmat, index=result_pid.index, columns=result_pid.columns) aniblastall_diff.to_csv(self.outdir / "pyani_aniblastall_diff.tab", sep="\t") self.assertLess(aniblastall_diff.abs().values.max(), self.tolerance["ANIblastall"])
def unified_anib(indirname,User_ID): # Build BLAST databases and run pairwise BLASTN # Fraglengths does not get reused with BLASTN os.mkdir(indirname+'{0}_out/'.format(User_ID)) os.system("chmod 777 {0}".format(indirname+'{0}_out'.format(User_ID))) logging.basicConfig(level=logging.DEBUG, filename="/home/linproject/Workspace/LIN_log/logfile_{0}".format(User_ID), filemode="a+", format="%(asctime)-15s %(levelname)-8s %(message)s") infiles = pyani_files.get_fasta_files(indirname) org_lengths = pyani_files.get_sequence_lengths(infiles) fragsize = pyani_config.FRAGSIZE filestems = pyani_config.ANIB_FILESTEMS filenames = os.listdir(indirname) for fname in filenames: if ' ' in os.path.abspath(fname): logging.error("File or directory '%s' contains whitespace" % fname) logging.error("This will cause issues with MUMmer and BLAST") logging.error("(exiting)") sys.exit(1) fragfiles, fraglengths = anib.fragment_FASTA_files(infiles, indirname+'{0}_out/'.format(User_ID), fragsize) # Export fragment lengths as JSON, in case we re-run BLASTALL with # --skip_blastn with open(os.path.join(indirname+'{0}_out/'.format(User_ID), 'fraglengths.json'), 'w') as outfile: json.dump(fraglengths, outfile) # Which executables are we using? format_exe = pyani_config.FORMATDB_DEFAULT blast_exe = pyani_config.BLASTALL_DEFAULT # Run BLAST database-building and executables from a jobgraph logging.info("Creating job dependency graph") jobgraph = anib.make_job_graph(infiles, fragfiles, indirname+'{0}_out/'.format(User_ID), format_exe, blast_exe, 'ANIblastall') logging.info("Running jobs with multiprocessing") logging.info("Running job dependency graph") cumval = run_mp.run_dependency_graph(jobgraph, verbose=False, logger=logging) if 0 < cumval: logging.warning("At least one BLAST run failed. " + "%s may fail." % 'ANIblastall') else: logging.info("All multiprocessing jobs complete.") # Process pairwise BLASTN output logging.info("Processing pairwise %s BLAST output." % 'ANIblastall') try: data = anib.process_blast(indirname+'{0}_out/'.format(User_ID), org_lengths, fraglengths=fraglengths, mode='ANIblastall') except ZeroDivisionError: logging.error("One or more BLAST output files has a problem.") if 0 < cumval: logging.error("This is possibly due to BLASTN run failure, " + "please investigate") else: logging.error("This is possibly due to ara BLASTN comparison " + "being too distant for use.") logging.error(last_exception()) return data[1]
def run_blast(args: Namespace, logger: Logger, infiles: List[Path], blastdir: Path) -> Tuple: """Run BLAST commands for ANIb methods. :param args: Namespace of command-line options :param logger: logging object :param infiles: iterable of sequence files to compare :param blastdir: path of directory to fragment BLASTN databases Runs BLAST database creation and comparisons, returning the cumulative return values of the BLAST tool subprocesses, and the fragment sizes for each input file """ if not args.skip_blastn: logger.info("Fragmenting input files, and writing to %s", args.outdirname) fragfiles, fraglengths = make_sequence_fragments( args, logger, infiles, blastdir) # Run BLAST database-building and executables from a jobgraph logger.info("Creating job dependency graph") jobgraph = anib.make_job_graph( infiles, fragfiles, anib.make_blastcmd_builder(args.method, blastdir)) if args.scheduler == "multiprocessing": logger.info("Running dependency graph with multiprocessing") cumval = run_mp.run_dependency_graph(jobgraph, logger=logger) if cumval > 0: logger.warning( f"At least one BLAST run failed. {args.method} may fail. Please investigate." ) else: logger.info("All multiprocessing jobs complete.") elif args.scheduler == "SGE": logger.info("Running dependency graph with SGE") run_sge.run_dependency_graph(jobgraph) else: logger.error( f"Scheduler {args.scheduler} not recognised (exiting)") raise SystemError(1) else: logger.warning("Skipping BLASTN runs (as instructed)!") # Import fragment lengths from JSON if args.method == "ANIblastall": fragpath = blastdir / "fraglengths.json" logger.info(f"Loading sequence fragments from {fragpath}") with open(fragpath, "rU") as ifh: fraglengths = json.load(ifh) else: fraglengths = dict() return cumval, fraglengths
def test_anib_concordance(self): """ANIb results concordant with JSpecies. We expect ANIb results to be quite different, as the BLASTN algorithm changed substantially between BLAST and BLAST+ """ # Perform ANIb on the input directory contents outdir = os.path.join(self.outdir, "blastn") os.makedirs(outdir, exist_ok=True) fragfiles, fraglengths = anib.fragment_fasta_files( self.infiles, outdir, self.fragsize) jobgraph = anib.make_job_graph( self.infiles, fragfiles, anib.make_blastcmd_builder("ANIb", outdir)) assert_equal(0, run_mp.run_dependency_graph(jobgraph)) results = anib.process_blast(outdir, self.orglengths, fraglengths, mode="ANIb") result_pid = results.percentage_identity result_pid.to_csv(os.path.join(self.outdir, "pyani_anib.tab"), sep="\t") # Compare JSpecies output to results. We do this in two blocks, # masked according to whether the expected result is greater than # 90% identity, or less than that threshold. # The complete difference matrix is written to output, though result_pid = result_pid.sort_index(axis=0).sort_index(axis=1) * 100.0 lo_result = result_pid.mask(result_pid >= 90).fillna(0) hi_result = result_pid.mask(result_pid < 90).fillna(0) lo_target = self.target["ANIb"].mask( self.target["ANIb"] >= 90).fillna(0) hi_target = self.target["ANIb"].mask( self.target["ANIb"] < 90).fillna(0) lo_diffmat = lo_result.as_matrix() - lo_target.as_matrix() hi_diffmat = hi_result.as_matrix() - hi_target.as_matrix() diffmat = result_pid.as_matrix() - self.target["ANIb"].as_matrix() lo_diff = pd.DataFrame(lo_diffmat, index=result_pid.index, columns=result_pid.columns) hi_diff = pd.DataFrame(hi_diffmat, index=result_pid.index, columns=result_pid.columns) anib_diff = pd.DataFrame(diffmat, index=result_pid.index, columns=result_pid.columns) anib_diff.to_csv(os.path.join(self.outdir, "pyani_anib_diff.tab"), sep="\t") assert_less(lo_diff.abs().values.max(), self.tolerance["ANIb_lo"]) assert_less(hi_diff.abs().values.max(), self.tolerance["ANIb_hi"])
def test_anib_concordance( paths_concordance_fna, path_concordance_jspecies, tolerance_anib_hi, tolerance_anib_lo, threshold_anib_lo_hi, fragment_length, tmp_path, ): """Check ANIb results are concordant with JSpecies. We expect ANIb results to be quite different, as the BLASTN algorithm changed substantially between BLAST and BLAST+ (the megaBLAST algorithm is now the default for BLASTN) """ # Get lengths of input genomes orglengths = pyani_files.get_sequence_lengths(paths_concordance_fna) # Build and run BLAST jobs fragfiles, fraglengths = anib.fragment_fasta_files( paths_concordance_fna, tmp_path, fragment_length ) jobgraph = anib.make_job_graph( paths_concordance_fna, fragfiles, anib.make_blastcmd_builder("ANIb", tmp_path) ) assert 0 == run_mp.run_dependency_graph(jobgraph) # Jobs must run correctly # Process BLAST output result_pid = anib.process_blast( tmp_path, orglengths, fraglengths, mode="ANIb" ).percentage_identity # Compare JSpecies output to results. We do this in two blocks, # masked according to whether the expected result is greater than # a threshold separating "low" from "high" identity comparisons. result_pid = result_pid.sort_index(axis=0).sort_index(axis=1) * 100.0 lo_result = result_pid.mask(result_pid >= threshold_anib_lo_hi).fillna(0).values hi_result = result_pid.mask(result_pid < threshold_anib_lo_hi).fillna(0).values tgt_pid = parse_jspecies(path_concordance_jspecies)["ANIb"] lo_target = tgt_pid.mask(tgt_pid >= threshold_anib_lo_hi).fillna(0).values hi_target = tgt_pid.mask(tgt_pid < threshold_anib_lo_hi).fillna(0).values assert (lo_result - lo_target, hi_result - hi_target) == ( pytest.approx(0, abs=tolerance_anib_lo), pytest.approx(0, abs=tolerance_anib_hi), )
def test_anib_concordance(self): """ANIb results concordant with JSpecies. We expect ANIb results to be quite different, as the BLASTN algorithm changed substantially between BLAST and BLAST+ """ # Perform ANIb on the input directory contents outdir = os.path.join(self.outdir, "blastn") os.makedirs(outdir, exist_ok=True) fragfiles, fraglengths = anib.fragment_fasta_files( self.infiles, outdir, self.fragsize ) jobgraph = anib.make_job_graph( self.infiles, fragfiles, anib.make_blastcmd_builder("ANIb", outdir) ) assert_equal(0, run_mp.run_dependency_graph(jobgraph)) results = anib.process_blast(outdir, self.orglengths, fraglengths, mode="ANIb") result_pid = results.percentage_identity result_pid.to_csv(os.path.join(self.outdir, "pyani_anib.tab"), sep="\t") # Compare JSpecies output to results. We do this in two blocks, # masked according to whether the expected result is greater than # 90% identity, or less than that threshold. # The complete difference matrix is written to output, though result_pid = result_pid.sort_index(axis=0).sort_index(axis=1) * 100.0 lo_result = result_pid.mask(result_pid >= 90).fillna(0) hi_result = result_pid.mask(result_pid < 90).fillna(0) lo_target = self.target["ANIb"].mask(self.target["ANIb"] >= 90).fillna(0) hi_target = self.target["ANIb"].mask(self.target["ANIb"] < 90).fillna(0) lo_diffmat = lo_result.values - lo_target.values hi_diffmat = hi_result.values - hi_target.values diffmat = result_pid.values - self.target["ANIb"].values lo_diff = pd.DataFrame( lo_diffmat, index=result_pid.index, columns=result_pid.columns ) hi_diff = pd.DataFrame( hi_diffmat, index=result_pid.index, columns=result_pid.columns ) anib_diff = pd.DataFrame( diffmat, index=result_pid.index, columns=result_pid.columns ) anib_diff.to_csv(os.path.join(self.outdir, "pyani_anib_diff.tab"), sep="\t") assert_less(lo_diff.abs().values.max(), self.tolerance["ANIb_lo"]) assert_less(hi_diff.abs().values.max(), self.tolerance["ANIb_hi"])
def calculate_anim(args: Namespace, infiles: List[Path], org_lengths: Dict) -> pyani_tools.ANIResults: """Return ANIm result dataframes for files in input directory. :param args: Namespace, command-line arguments :param logger: logging object :param infiles: list of paths to each input file :param org_lengths: dict, input sequence lengths, keyed by sequence Finds ANI by the ANIm method, as described in Richter et al (2009) Proc Natl Acad Sci USA 106: 19126-19131 doi:10.1073/pnas.0906412106. All FASTA format files (selected by suffix) in the input directory are compared against each other, pairwise, using NUCmer (which must be in the path). NUCmer output is stored in the output directory. The NUCmer .delta file output is parsed to obtain an alignment length and similarity error count for every unique region alignment between the two organisms, as represented by the sequences in the FASTA files. These are processed to give matrices of aligned sequence lengths, average nucleotide identity (ANI) percentages, coverage (aligned percentage of whole genome), and similarity error cound for each pairwise comparison. """ logger = logging.getLogger(__name__) logger.info("Running ANIm") logger.info("Generating NUCmer command-lines") deltadir = args.outdirname / ALIGNDIR["ANIm"] logger.info("Writing nucmer output to %s", deltadir) # Schedule NUCmer runs if not args.skip_nucmer: joblist = anim.generate_nucmer_jobs( infiles, args.outdirname, nucmer_exe=args.nucmer_exe, filter_exe=args.filter_exe, maxmatch=args.maxmatch, jobprefix=args.jobprefix, ) if args.scheduler == "multiprocessing": logger.info("Running jobs with multiprocessing") if args.workers is None: logger.info( "(using maximum number of available worker threads)") else: logger.info("(using %d worker threads, if available)", args.workers) cumval = run_mp.run_dependency_graph(joblist, workers=args.workers, logger=logger) logger.info("Cumulative return value: %d", cumval) if cumval > 0: logger.warning( "At least one NUCmer comparison failed. ANIm may fail.") else: logger.info("All multiprocessing jobs complete.") else: logger.info("Running jobs with SGE") logger.info("Jobarray group size set to %d", args.sgegroupsize) run_sge.run_dependency_graph( joblist, jgprefix=args.jobprefix, sgegroupsize=args.sgegroupsize, sgeargs=args.sgeargs, ) else: logger.warning("Skipping NUCmer run (as instructed)!") # Process resulting .delta files logger.info("Processing NUCmer .delta files.") results = anim.process_deltadir(deltadir, org_lengths, logger=logger) if results.zero_error: # zero percentage identity error if not args.skip_nucmer and args.scheduler == "multiprocessing": if cumval > 0: logger.error( "This has possibly been a NUCmer run failure, please investigate", exc_info=True, ) raise SystemExit(1) logger.error( "This is possibly due to:\n\t(i) a NUCmer comparison being too distant " "for use (please consider using the --maxmatch option)\n\t(ii) NUCmer run " "failure (analysis will continue, but please investigate)") if not args.nocompress: logger.info("Compressing/deleting %s", deltadir) compress_delete_outdir(deltadir, logger) # Return processed data from .delta files return results
def unified_anib(infiles, org_lengths): """Calculate ANIb for files in input directory. - infiles - paths to each input file - org_lengths - dictionary of input sequence lengths, keyed by sequence Calculates ANI by the ANIb method, as described in Goris et al. (2007) Int J Syst Evol Micr 57: 81-91. doi:10.1099/ijs.0.64483-0. There are some minor differences depending on whether BLAST+ or legacy BLAST (BLASTALL) methods are used. All FASTA format files (selected by suffix) in the input directory are used to construct BLAST databases, placed in the output directory. Each file's contents are also split into sequence fragments of length options.fragsize, and the multiple FASTA file that results written to the output directory. These are BLASTNed, pairwise, against the databases. The BLAST output is interrogated for all fragment matches that cover at least 70% of the query sequence, with at least 30% nucleotide identity over the full length of the query sequence. This is an odd choice and doesn't correspond to the twilight zone limit as implied by Goris et al. We persist with their definition, however. Only these qualifying matches contribute to the total aligned length, and total aligned sequence identity used to calculate ANI. The results are processed to give matrices of aligned sequence length (aln_lengths.tab), similarity error counts (sim_errors.tab), ANIs (perc_ids.tab), and minimum aligned percentage (perc_aln.tab) of each genome, for each pairwise comparison. These are written to the output directory in plain text tab-separated format. """ logger.info("Running %s", args.method) blastdir = os.path.join(args.outdirname, ALIGNDIR[args.method]) logger.info("Writing BLAST output to %s", blastdir) # Build BLAST databases and run pairwise BLASTN if not args.skip_blastn: # Make sequence fragments logger.info("Fragmenting input files, and writing to %s", args.outdirname) # Fraglengths does not get reused with BLASTN fragfiles, fraglengths = anib.fragment_fasta_files( infiles, blastdir, args.fragsize) # Export fragment lengths as JSON, in case we re-run with --skip_blastn with open(os.path.join(blastdir, 'fraglengths.json'), 'w') as outfile: json.dump(fraglengths, outfile) # Which executables are we using? #if args.method == "ANIblastall": # format_exe = args.formatdb_exe # blast_exe = args.blastall_exe #else: # format_exe = args.makeblastdb_exe # blast_exe = args.blastn_exe # Run BLAST database-building and executables from a jobgraph logger.info("Creating job dependency graph") jobgraph = anib.make_job_graph( infiles, fragfiles, anib.make_blastcmd_builder(args.method, blastdir)) #jobgraph = anib.make_job_graph(infiles, fragfiles, blastdir, # format_exe, blast_exe, args.method, # jobprefix=args.jobprefix) if args.scheduler == 'multiprocessing': logger.info("Running jobs with multiprocessing") logger.info("Running job dependency graph") cumval = run_mp.run_dependency_graph(jobgraph, logger=logger) if 0 < cumval: logger.warning( "At least one BLAST run failed. " + "%s may fail.", args.method) else: logger.info("All multiprocessing jobs complete.") else: run_sge.run_dependency_graph(jobgraph, logger=logger) logger.info("Running jobs with SGE") else: # Import fragment lengths from JSON if args.method == "ANIblastall": with open(os.path.join(blastdir, 'fraglengths.json'), 'rU') as infile: fraglengths = json.load(infile) else: fraglengths = None logger.warning("Skipping BLASTN runs (as instructed)!") # Process pairwise BLASTN output logger.info("Processing pairwise %s BLAST output.", args.method) try: data = anib.process_blast(blastdir, org_lengths, fraglengths=fraglengths, mode=args.method) except ZeroDivisionError: logger.error("One or more BLAST output files has a problem.") if not args.skip_blastn: if 0 < cumval: logger.error("This is possibly due to BLASTN run failure, " + "please investigate") else: logger.error("This is possibly due to a BLASTN comparison " + "being too distant for use.") logger.error(last_exception()) if not args.nocompress: logger.info("Compressing/deleting %s", blastdir) compress_delete_outdir(blastdir) # Return processed BLAST data return data
def calculate_anim(infiles, org_lengths): """Returns ANIm result dataframes for files in input directory. - infiles - paths to each input file - org_lengths - dictionary of input sequence lengths, keyed by sequence Finds ANI by the ANIm method, as described in Richter et al (2009) Proc Natl Acad Sci USA 106: 19126-19131 doi:10.1073/pnas.0906412106. All FASTA format files (selected by suffix) in the input directory are compared against each other, pairwise, using NUCmer (which must be in the path). NUCmer output is stored in the output directory. The NUCmer .delta file output is parsed to obtain an alignment length and similarity error count for every unique region alignment between the two organisms, as represented by the sequences in the FASTA files. These are processed to give matrices of aligned sequence lengths, average nucleotide identity (ANI) percentages, coverage (aligned percentage of whole genome), and similarity error cound for each pairwise comparison. """ logger.info("Running ANIm") logger.info("Generating NUCmer command-lines") deltadir = os.path.join(args.outdirname, ALIGNDIR['ANIm']) logger.info("Writing nucmer output to %s", deltadir) # Schedule NUCmer runs if not args.skip_nucmer: joblist = anim.generate_nucmer_jobs(infiles, args.outdirname, nucmer_exe=args.nucmer_exe, jobprefix=args.jobprefix) if args.scheduler == 'multiprocessing': logger.info("Running jobs with multiprocessing") if args.workers is None: logger.info("(using maximum number of available " + "worker threads)") else: logger.info("(using %d worker threads, if available)", args.workers) cumval = run_mp.run_dependency_graph(joblist, workers=args.workers, logger=logger) logger.info("Cumulative return value: %d", cumval) if 0 < cumval: logger.warning("At least one NUCmer comparison failed. " + "ANIm may fail.") else: logger.info("All multiprocessing jobs complete.") else: logger.info("Running jobs with SGE") logger.info("Jobarray group size set to %d", args.sgegroupsize) run_sge.run_dependency_graph(joblist, logger=logger, jgprefix=args.jobprefix, sgegroupsize=args.sgegroupsize) else: logger.warning("Skipping NUCmer run (as instructed)!") # Process resulting .delta files logger.info("Processing NUCmer .delta files.") results = anim.process_deltadir(deltadir, org_lengths, logger=logger) if results.zero_error: # zero percentage identity error if not args.skip_nucmer and args.scheduler == 'multiprocessing': if 0 < cumval: logger.error("This has possibly been a NUCmer run failure, " + "please investigate") logger.error(last_exception()) sys.exit(1) else: logger.error("This is possibly due to a NUCmer comparison " + "being too distant for use. Please consider " + "using the --maxmatch option.") logger.error("This is alternatively due to NUCmer run " + "failure, analysis will continue, but please " + "investigate.") if not args.nocompress: logger.info("Compressing/deleting %s", deltadir) compress_delete_outdir(deltadir) # Return processed data from .delta files return results
def calculate_anim(infiles, org_lengths): """Returns ANIm result dataframes for files in input directory. - infiles - paths to each input file - org_lengths - dictionary of input sequence lengths, keyed by sequence Finds ANI by the ANIm method, as described in Richter et al (2009) Proc Natl Acad Sci USA 106: 19126-19131 doi:10.1073/pnas.0906412106. All FASTA format files (selected by suffix) in the input directory are compared against each other, pairwise, using NUCmer (which must be in the path). NUCmer output is stored in the output directory. The NUCmer .delta file output is parsed to obtain an alignment length and similarity error count for every unique region alignment between the two organisms, as represented by the sequences in the FASTA files. These are processed to give matrices of aligned sequence lengths, average nucleotide identity (ANI) percentages, coverage (aligned percentage of whole genome), and similarity error cound for each pairwise comparison. """ logger.info("Running ANIm") logger.info("Generating NUCmer command-lines") deltadir = os.path.join(args.outdirname, ALIGNDIR['ANIm']) logger.info("Writing nucmer output to %s" % deltadir) # Schedule NUCmer runs if not args.skip_nucmer: joblist = anim.generate_nucmer_jobs(infiles, args.outdirname, nucmer_exe=args.nucmer_exe, maxmatch=args.maxmatch, jobprefix=args.jobprefix) if args.scheduler == 'multiprocessing': logger.info("Running jobs with multiprocessing") if args.workers is None: logger.info("(using maximum number of available " + "worker threads)") else: logger.info("(using %d worker threads, if available)" % args.workers) cumval = run_mp.run_dependency_graph(joblist, workers=args.workers, verbose=args.verbose, logger=logger) logger.info("Cumulative return value: %d" % cumval) if 0 < cumval: logger.warning("At least one NUCmer comparison failed. " + "ANIm may fail.") else: logger.info("All multiprocessing jobs complete.") else: logger.info("Running jobs with SGE") logger.info("Jobarray group size set to %d" % args.sgegroupsize) run_sge.run_dependency_graph(joblist, verbose=args.verbose, logger=logger, jgprefix=args.jobprefix, sgegroupsize=args.sgegroupsize) else: logger.warning("Skipping NUCmer run (as instructed)!") # Process resulting .delta files logger.info("Processing NUCmer .delta files.") data = anim.process_deltadir(deltadir, org_lengths, logger=logger) if data[-1]: # zero percentage identity error if not args.skip_nucmer and args.scheduler == 'multiprocessing': if 0 < cumval: logger.error("This has possibly been a NUCmer run failure, " + "please investigate") logger.error(last_exception()) sys.exit(1) else: logger.error("This is possibly due to a NUCmer comparison " + "being too distant for use. Please consider " + "using the --maxmatch option.") logger.error("This is alternatively due to NUCmer run " + "failure, analysis will continue, but please " + "investigate.") if not args.nocompress: logger.info("Compressing/deleting %s" % deltadir) compress_delete_outdir(deltadir) # Return processed data from .delta files return tuple(data[:-1])
def unified_anib(infiles, org_lengths): """Calculate ANIb for files in input directory. - infiles - paths to each input file - org_lengths - dictionary of input sequence lengths, keyed by sequence Calculates ANI by the ANIb method, as described in Goris et al. (2007) Int J Syst Evol Micr 57: 81-91. doi:10.1099/ijs.0.64483-0. There are some minor differences depending on whether BLAST+ or legacy BLAST (BLASTALL) methods are used. All FASTA format files (selected by suffix) in the input directory are used to construct BLAST databases, placed in the output directory. Each file's contents are also split into sequence fragments of length options.fragsize, and the multiple FASTA file that results written to the output directory. These are BLASTNed, pairwise, against the databases. The BLAST output is interrogated for all fragment matches that cover at least 70% of the query sequence, with at least 30% nucleotide identity over the full length of the query sequence. This is an odd choice and doesn't correspond to the twilight zone limit as implied by Goris et al. We persist with their definition, however. Only these qualifying matches contribute to the total aligned length, and total aligned sequence identity used to calculate ANI. The results are processed to give matrices of aligned sequence length (aln_lengths.tab), similarity error counts (sim_errors.tab), ANIs (perc_ids.tab), and minimum aligned percentage (perc_aln.tab) of each genome, for each pairwise comparison. These are written to the output directory in plain text tab-separated format. """ logger.info("Running %s" % args.method) # Build BLAST databases and run pairwise BLASTN if not args.skip_blastn: # Make sequence fragments logger.info("Fragmenting input files, and writing to %s" % args.outdirname) # Fraglengths does not get reused with BLASTN fragfiles, fraglengths = anib.fragment_FASTA_files(infiles, args.outdirname, args.fragsize) # Export fragment lengths as JSON, in case we re-run BLASTALL with # --skip_blastn if args.method == "ANIblastall": with open(os.path.join(args.outdirname, 'fraglengths.json'), 'w') as outfile: json.dump(fraglengths, outfile) # Which executables are we using? if args.method == "ANIblastall": format_exe = args.formatdb_exe blast_exe = args.blastall_exe else: format_exe = args.makeblastdb_exe blast_exe = args.blastn_exe # Run BLAST database-building and executables from a jobgraph logger.info("Creating job dependency graph") jobgraph = anib.make_job_graph(infiles, fragfiles, args.outdirname, format_exe, blast_exe, args.method) if args.scheduler == 'multiprocessing': logger.info("Running jobs with multiprocessing") logger.info("Running job dependency graph") cumval = run_mp.run_dependency_graph(jobgraph, verbose=args.verbose, logger=logger) if 0 < cumval: logger.warning("At least one BLAST run failed. " + "%s may fail." % args.method) else: logger.info("All multiprocessing jobs complete.") else: run_sge.run_dependency_graph(jobgraph, verbose=args.verbose, logger=logger) logger.info("Running jobs with SGE") else: # Import fragment lengths from JSON if args.method == "ANIblastall": with open(os.path.join(args.outdirname, 'fraglengths.json'), 'rU') as infile: fraglengths = json.load(infile) else: fraglengths = None logger.warning("Skipping BLASTN runs (as instructed)!") # Process pairwise BLASTN output logger.info("Processing pairwise %s BLAST output." % args.method) try: data = anib.process_blast(args.outdirname, org_lengths, fraglengths=fraglengths, mode=args.method) except ZeroDivisionError: logger.error("One or more BLAST output files has a problem.") if not args.skip_blastn: if 0 < cumval: logger.error("This is possibly due to BLASTN run failure, " + "please investigate") else: logger.error("This is possibly due to a BLASTN comparison " + "being too distant for use.") logger.error(last_exception()) return data
def calculate_anim(infiles, org_lengths): """Returns ANIm result dataframes for files in input directory. - infiles - paths to each input file - org_lengths - dictionary of input sequence lengths, keyed by sequence Finds ANI by the ANIm method, as described in Richter et al (2009) Proc Natl Acad Sci USA 106: 19126-19131 doi:10.1073/pnas.0906412106. All FASTA format files (selected by suffix) in the input directory are compared against each other, pairwise, using NUCmer (which must be in the path). NUCmer output is stored in the output directory. The NUCmer .delta file output is parsed to obtain an alignment length and similarity error count for every unique region alignment between the two organisms, as represented by the sequences in the FASTA files. These are processed to give matrices of aligned sequence lengths, average nucleotide identity (ANI) percentages, coverage (aligned percentage of whole genome), and similarity error cound for each pairwise comparison. """ logger.info("Running ANIm") logger.info("Generating NUCmer command-lines") # Schedule NUCmer runs if not args.skip_nucmer: joblist = anim.generate_nucmer_jobs(infiles, args.outdirname, nucmer_exe=args.nucmer_exe, maxmatch=args.maxmatch) if args.scheduler == 'multiprocessing': logger.info("Running jobs with multiprocessing") cumval = run_mp.run_dependency_graph(joblist, verbose=args.verbose, logger=logger) logger.info("Cumulative return value: %d" % cumval) if 0 < cumval: logger.warning("At least one NUCmer comparison failed. " + "ANIm may fail.") else: logger.info("All multiprocessing jobs complete.") else: logger.info("Running jobs with SGE") run_sge.run_dependency_graph(joblist, verbose=args.verbose, logger=logger) else: logger.warning("Skipping NUCmer run (as instructed)!") # Process resulting .delta files logger.info("Processing NUCmer .delta files.") try: data = anim.process_deltadir(args.outdirname, org_lengths) except ZeroDivisionError: logger.error("One or more NUCmer output files has a problem.") if not args.skip_nucmer: if 0 < cumval: logger.error("This is possibly due to NUCmer run failure, " + "please investigate") else: logger.error("This is possibly due to a NUCmer comparison " + "being too distant for use. Please consider " + "using the --maxmatch option.") logger.error(last_exception()) return data