public void OneTranscriptOneHeterozygousSynonymous()
{
Genome genome = new Genome(Path.Combine(TestContext.CurrentContext.TestDirectory, "TestData", "chr1_sample.fa"));
VCFParser vcf = new VCFParser(Path.Combine(TestContext.CurrentContext.TestDirectory, "TestVcfs", "chr_1_one_heterozygous_synonymous.vcf"));
List <Variant> variants = vcf.Select(x => new Variant(null, x, genome)).ToList();
Assert.AreEqual(1, variants.Count);
GeneModel geneModel = new GeneModel(genome, Path.Combine(TestContext.CurrentContext.TestDirectory, "TestData", "chr1_one_transcript.gtf"));
List <Protein> proteins_wo_variant = geneModel.Translate(true).ToList();
List <Transcript> transcripts = geneModel.ApplyVariants(variants);
List <Protein> proteins = transcripts.Select(t => t.Protein()).ToList();
Assert.AreEqual(1, geneModel.Genes.Count);
Assert.AreEqual(1, proteins.Count);
Assert.AreEqual(1, proteins_wo_variant.Count);
Assert.AreEqual(1, new HashSet <string> {
proteins[0].BaseSequence, proteins_wo_variant[0].BaseSequence
}.Count);
Assert.IsTrue(proteins.Any(p => p.FullName.Contains(FunctionalClass.SILENT.ToString()))); // synonymous
Assert.IsTrue(proteins.Any(p => p.FullName.Contains(GenotypeType.HETEROZYGOUS.ToString()))); // synonymous
Assert.IsTrue(proteins.Any(p => p.FullName.Contains("1:69666")));
string proteinFasta = Path.Combine(TestContext.CurrentContext.TestDirectory, "TestVcfs", "chr_1_one_heterozygous_synonymous.fasta");
ProteinDbWriter.WriteFastaDatabase(proteins, proteinFasta, " ");
string[] proteinFastaLines = File.ReadLines(proteinFasta).ToArray();
Assert.IsTrue(proteinFastaLines[0].Contains(FunctionalClass.SILENT.ToString())); // synonymous
Assert.IsTrue(proteinFastaLines[0].Contains("1:69666"));
}
public void Test_read_xml_write_read_fasta()
{
ModificationMotif.TryGetMotif("X", out ModificationMotif motif);
var nice = new List <Modification>
{
new Modification("fayk", null, "mt", null, motif, "Anywhere.", null, null, null, null, null, null, null, null)
};
List <Protein> ok = ProteinDbLoader.LoadProteinXML(Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"xml2.xml"), true, DecoyType.None, nice, false, null,
out Dictionary <string, Modification> un);
ProteinDbWriter.WriteFastaDatabase(ok, Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"rewrite_xml_test.fasta"), "|");
List <Protein> ok2 = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"rewrite_xml_test.fasta"), true, DecoyType.None, false,
ProteinDbLoader.UniprotAccessionRegex, ProteinDbLoader.UniprotFullNameRegex, ProteinDbLoader.UniprotNameRegex, ProteinDbLoader.UniprotGeneNameRegex, ProteinDbLoader.UniprotOrganismRegex, out var b);
Assert.AreEqual(ok.Count, ok2.Count);
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].BaseSequence == ok2[i].BaseSequence));
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].Name == ok2[i].Name));
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].Organism == ok2[i].Organism));
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].GeneNames.First().Item2 == ok2[i].GeneNames.First().Item2));
Assert.True(ok.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedBeginPosition == null || prod.OneBasedBeginPosition > 0 && prod.OneBasedBeginPosition <= p.Length)));
Assert.True(ok.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedEndPosition == null || prod.OneBasedEndPosition > 0 && prod.OneBasedEndPosition <= p.Length)));
Assert.True(ok2.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedBeginPosition == null || prod.OneBasedBeginPosition > 0 && prod.OneBasedBeginPosition <= p.Length)));
Assert.True(ok2.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedEndPosition == null || prod.OneBasedEndPosition > 0 && prod.OneBasedEndPosition <= p.Length)));
}
public static void FastaTest()
{
List <Protein> prots = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"fasta.fasta"), true, DecoyType.Reverse, false,
ProteinDbLoader.UniprotAccessionRegex, ProteinDbLoader.UniprotFullNameRegex, ProteinDbLoader.UniprotNameRegex, ProteinDbLoader.UniprotGeneNameRegex,
ProteinDbLoader.UniprotOrganismRegex, out var a);
ProteinDbWriter.WriteFastaDatabase(prots, Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"rewrite_fasta.fasta"), "|");
List <Protein> prots2 = ProteinDbLoader.LoadProteinFasta(
Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"rewrite_fasta.fasta"),
true,
DecoyType.None,
false,
ProteinDbLoader.UniprotAccessionRegex, ProteinDbLoader.UniprotFullNameRegex, ProteinDbLoader.UniprotNameRegex, ProteinDbLoader.UniprotGeneNameRegex,
ProteinDbLoader.UniprotOrganismRegex,
out var un);
Assert.AreEqual("P62805", prots.First().Accession);
Assert.AreEqual("H4_HUMAN", prots.First().Name);
Assert.AreEqual("Histone H4", prots.First().FullName);
Assert.AreEqual("HIST1H4A", prots.First().GeneNames.First().Item2);
Assert.AreEqual("H**o sapiens", prots.First().Organism);
Assert.AreEqual("P62805", prots2.First().Accession);
Assert.AreEqual("H4_HUMAN", prots2.First().Name);
Assert.AreEqual("Histone H4", prots2.First().FullName);
Assert.AreEqual("HIST1H4A", prots2.First().GeneNames.First().Item2);
Assert.AreEqual("H**o sapiens", prots2.First().Organism);
}
public List <string> TransferModifications(string spritzDirectory, string sourceXmlPath, List <string> destinationXmlPaths, List <Protein> additionalProteins)
{
var uniprotPtms = ProteinAnnotation.GetUniProtMods(spritzDirectory);
List <string> outxmls = new List <string>();
var uniprot = File.Exists(sourceXmlPath) ?
ProteinDbLoader.LoadProteinXML(sourceXmlPath, true, DecoyType.None, uniprotPtms, false, null, out Dictionary <string, Modification> un) :
new List <Protein>();
foreach (var xml in destinationXmlPaths)
{
if (xml == null || !File.Exists(xml))
{
continue;
}
string outxml = Path.Combine(Path.GetDirectoryName(xml), Path.GetFileNameWithoutExtension(xml) + ".withmods.xml");
var nonVariantProts = ProteinDbLoader.LoadProteinXML(xml, true, DecoyType.None, uniprotPtms, false, null, out un).Select(p => p.NonVariantProtein).Distinct();
var newProts = ProteinAnnotation.CombineAndAnnotateProteins(uniprot, nonVariantProts.Concat(additionalProteins).ToList());
ProteinDbWriter.WriteXmlDatabase(null, newProts, outxml);
string outfasta = Path.Combine(Path.GetDirectoryName(xml), Path.GetFileNameWithoutExtension(xml) + ".spritz.fasta");
ProteinDbWriter.WriteFastaDatabase(newProts.SelectMany(p => p.GetVariantProteins()).ToList(), outfasta, "|");
outxmls.Add(outxml);
}
return(outxmls);
}
public void Test_accession_regex_weird()
{
Regex bad = new Regex(@"/()/");
List <Protein> ok = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, @"test_ensembl.pep.all.fasta"), false, false, bad, bad, bad, bad);
ProteinDbWriter.WriteFastaDatabase(ok, Path.Combine(TestContext.CurrentContext.TestDirectory, @"rewrite_test_ensembl.pep.all.fasta"), " ");
List <Protein> ok2 = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, @"rewrite_test_ensembl.pep.all.fasta"), false, false, bad, bad, bad, bad);
Assert.AreEqual("ENSP00000381386 pep:known chromosome:GRCh37:22:24313554:24316773:-1 gene:ENSG00000099977 transcript:ENST00000398344 gene_biotype:protein_coding transcript_biotype:protein_coding", ok[0].Accession);
Assert.AreEqual("ENSP00000381386 pep:known chromosome:GRCh37:22:24313554:24316773:-1 gene:ENSG00000099977 transcript:ENST00000398344 gene_biotype:protein_coding transcript_biotype:protein_coding", ok2[0].Accession);
Assert.AreEqual(ok.Count, ok2.Count);
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].BaseSequence == ok2[i].BaseSequence));
}
public void MakeVariantFasta()
{
List <Protein> proteins = ProteinDbLoader.LoadProteinXML(@"E:\ProjectsActive\JurkatProteogenomics\180413\combined_1-trimmed-pair1Aligned.sortedByCoord.outProcessed.out.fixedQuals.split.concat.sorted.snpEffAnnotated.protein.xml",
true, DecoyType.None, null, false, null, out var un);
List <ProteinWithAppliedVariants> variantProteins = proteins.SelectMany(p => p.GetVariantProteins()).ToList();
List <ProteinWithAppliedVariants> variantProteins2 = variantProteins.Select(p =>
new ProteinWithAppliedVariants(
p.BaseSequence,
new Protein(p.Protein.BaseSequence, p.Protein.Accession, p.Protein.Organism, p.Protein.GeneNames.ToList(), p.Protein.OneBasedPossibleLocalizedModifications, p.Protein.ProteolysisProducts.ToList(), p.Protein.Name + string.Join(",", p.AppliedSequenceVariations.Select(d => d.Description)), p.Protein.FullName + string.Join(",", p.AppliedSequenceVariations.Select(d => d.Description)), p.IsDecoy, p.IsContaminant, p.DatabaseReferences.ToList(), p.SequenceVariations.ToList(), p.DisulfideBonds.ToList(), p.DatabaseFilePath),
p.AppliedSequenceVariations, p.Individual)
).ToList();
ProteinDbWriter.WriteFastaDatabase(variantProteins2.OfType <Protein>().ToList(), @"E:\ProjectsActive\Spritz\mmTesting\variantproteins.fasta", "|");
}
public void Test_read_write_read_fasta()
{
List <Protein> ok = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, @"test_ensembl.pep.all.fasta"), false, false, ProteinDbLoader.ensembl_accession_expression, ProteinDbLoader.ensembl_fullName_expression, ProteinDbLoader.ensembl_accession_expression, ProteinDbLoader.ensembl_gene_expression);
ProteinDbWriter.WriteFastaDatabase(ok, Path.Combine(TestContext.CurrentContext.TestDirectory, @"rewrite_test_ensembl.pep.all.fasta"), " ");
List <Protein> ok2 = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, @"rewrite_test_ensembl.pep.all.fasta"), false, false, ProteinDbLoader.ensembl_accession_expression, ProteinDbLoader.ensembl_fullName_expression, ProteinDbLoader.ensembl_accession_expression, ProteinDbLoader.ensembl_gene_expression);
Assert.AreEqual(ok.Count, ok2.Count);
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].BaseSequence == ok2[i].BaseSequence));
Assert.True(ok.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedBeginPosition == null || prod.OneBasedBeginPosition > 0 && prod.OneBasedBeginPosition <= p.Length)));
Assert.True(ok.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedEndPosition == null || prod.OneBasedEndPosition > 0 && prod.OneBasedEndPosition <= p.Length)));
Assert.True(ok2.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedBeginPosition == null || prod.OneBasedBeginPosition > 0 && prod.OneBasedBeginPosition <= p.Length)));
Assert.True(ok2.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedEndPosition == null || prod.OneBasedEndPosition > 0 && prod.OneBasedEndPosition <= p.Length)));
}
public static string TransferModifications(string sourceXmlPath, string destinationXmlPath)
{
var uniprotPtms = ProteinAnnotation.GetUniProtMods(Environment.CurrentDirectory);
var uniprot = ProteinDbLoader.LoadProteinXML(sourceXmlPath, true, DecoyType.None, uniprotPtms, false, null, out var un);
string outxml = Path.Combine(Path.GetDirectoryName(destinationXmlPath), Path.GetFileNameWithoutExtension(destinationXmlPath) + ".withmods.xml");
var nonVariantProts = ProteinDbLoader.LoadProteinXML(destinationXmlPath, true, DecoyType.None, uniprotPtms, false, null, out un).Select(p => p.NonVariantProtein).Distinct();
var newProts = ProteinAnnotation.CombineAndAnnotateProteins(uniprot, nonVariantProts.ToList());
ProteinDbWriter.WriteXmlDatabase(null, newProts, outxml);
string outfasta = Path.Combine(Path.GetDirectoryName(destinationXmlPath), Path.GetFileNameWithoutExtension(destinationXmlPath) + ".fasta");
var prot = newProts.FirstOrDefault(p => p.Accession.Contains("_"));
ProteinDbWriter.WriteFastaDatabase(newProts.SelectMany(p => p.GetVariantProteins()).ToList(), outfasta, "|");
return(outxml);
}
public void Test_read_write_read_fasta()
{
List <Protein> ok = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"test_ensembl.pep.all.fasta"), true, DecoyType.None, false,
ProteinDbLoader.EnsemblAccessionRegex, ProteinDbLoader.EnsemblFullNameRegex, ProteinDbLoader.EnsemblAccessionRegex, ProteinDbLoader.EnsemblGeneNameRegex, null, out var a);
ProteinDbWriter.WriteFastaDatabase(ok, Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"rewrite_test_ensembl.pep.all.fasta"), " ");
List <Protein> ok2 = ProteinDbLoader.LoadProteinFasta(Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"rewrite_test_ensembl.pep.all.fasta"), true, DecoyType.None, false,
ProteinDbLoader.EnsemblAccessionRegex, ProteinDbLoader.EnsemblFullNameRegex, ProteinDbLoader.EnsemblAccessionRegex, ProteinDbLoader.EnsemblGeneNameRegex, null, out var b);
Assert.AreEqual(ok.Count, ok2.Count);
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].BaseSequence == ok2[i].BaseSequence));
Assert.True(ok.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedBeginPosition == null || prod.OneBasedBeginPosition > 0 && prod.OneBasedBeginPosition <= p.Length)));
Assert.True(ok.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedEndPosition == null || prod.OneBasedEndPosition > 0 && prod.OneBasedEndPosition <= p.Length)));
Assert.True(ok2.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedBeginPosition == null || prod.OneBasedBeginPosition > 0 && prod.OneBasedBeginPosition <= p.Length)));
Assert.True(ok2.All(p => p.ProteolysisProducts.All(prod => prod.OneBasedEndPosition == null || prod.OneBasedEndPosition > 0 && prod.OneBasedEndPosition <= p.Length)));
}
public static void TestStringSanitation()
{
string messedUpSequence = @"PRO�EIN�";
// just test the string sanitation method alone
var sanitized = ProteinDbLoader.SanitizeAminoAcidSequence(messedUpSequence, 'X');
Assert.That(sanitized == "PROXEINX");
// test reading from a fasta
Protein protein = new Protein(messedUpSequence, "accession");
string fastaPath = Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"messedUp.fasta");
ProteinDbWriter.WriteFastaDatabase(new List <Protein> {
protein
}, fastaPath, "|");
var fastaProteins = ProteinDbLoader.LoadProteinFasta(fastaPath, true, DecoyType.Reverse, false, ProteinDbLoader.UniprotAccessionRegex,
ProteinDbLoader.UniprotFullNameRegex, ProteinDbLoader.UniprotNameRegex, ProteinDbLoader.UniprotGeneNameRegex,
ProteinDbLoader.UniprotOrganismRegex, out var a);
Assert.That(fastaProteins.First(p => !p.IsDecoy).BaseSequence == "PROXEINX");
// digest and fragment to check that there isn't a crash
var peptides = fastaProteins.First().Digest(new DigestionParams(), new List <Modification>(), new List <Modification>()).ToList();
foreach (PeptideWithSetModifications peptide in peptides)
{
List <Product> fragments = new List <Product>();
peptide.Fragment(DissociationType.HCD, FragmentationTerminus.Both, fragments);
}
// test reading from an XML
string xmlPath = Path.Combine(TestContext.CurrentContext.TestDirectory, "DatabaseTests", @"messedUp.xml");
ProteinDbWriter.WriteXmlDatabase(new Dictionary <string, HashSet <Tuple <int, Modification> > >(), new List <Protein> {
protein
}, xmlPath);
var xmlProteins = ProteinDbLoader.LoadProteinXML(xmlPath, true, DecoyType.Reverse, new List <Modification>(), false, new List <string>(), out var unk);
Assert.That(xmlProteins.First(p => !p.IsDecoy).BaseSequence == "PROXEINX");
}
public void test_read_write_read_fasta()
{
var nice = new List <Modification>
{
new ModificationWithLocation("fayk", null, null, ModificationSites.A, null, null)
};
Dictionary <string, Modification> un;
List <Protein> ok = ProteinDbLoader.LoadProteinDb(Path.Combine(TestContext.CurrentContext.TestDirectory, @"test_ensembl.pep.all.fasta"), false, nice, false, out un);
ProteinDbWriter.WriteFastaDatabase(ok, Path.Combine(TestContext.CurrentContext.TestDirectory, @"rewrite_test_ensembl.pep.all.fasta"));
List <Protein> ok2 = ProteinDbLoader.LoadProteinDb(Path.Combine(TestContext.CurrentContext.TestDirectory, @"rewrite_test_ensembl.pep.all.fasta"), false, nice, false, out un);
Assert.AreEqual(ok.Count, ok2.Count);
Assert.True(Enumerable.Range(0, ok.Count).All(i => ok[i].BaseSequence == ok2[i].BaseSequence));
Assert.True(ok.All(p => p.OneBasedBeginPositions.All(begin => begin == null || begin > 0 && begin <= p.Length)));
Assert.True(ok.All(p => p.OneBasedEndPositions.All(end => end == null || end > 0 && end <= p.Length)));
Assert.True(ok2.All(p => p.OneBasedBeginPositions.All(begin => begin == null || begin > 0 && begin <= p.Length)));
Assert.True(ok2.All(p => p.OneBasedEndPositions.All(end => end == null || end > 0 && end <= p.Length)));
}
public static string TransferModifications(string sourceXmlPath, string destinationXmlPath)
{
var uniprotPtms = ProteinAnnotation.GetUniProtMods(Environment.CurrentDirectory);
var uniprot = ProteinDbLoader.LoadProteinXML(sourceXmlPath, true, DecoyType.None, uniprotPtms, false, null, out var un);
string outxml = Path.Combine(Path.GetDirectoryName(destinationXmlPath), Path.GetFileNameWithoutExtension(destinationXmlPath) + ".withmods.xml");
var nonVariantProts = destinationXmlPath.EndsWith(".xml") | destinationXmlPath.EndsWith(".xml.gz") ?
ProteinDbLoader.LoadProteinXML(destinationXmlPath, true, DecoyType.None, uniprotPtms, false, null, out un).Select(p => p.NonVariantProtein).Distinct() :
ProteinDbLoader.LoadProteinFasta(destinationXmlPath, true, DecoyType.None, false, ProteinDbLoader.UniprotAccessionRegex, PgmNameRegex, PgmNameRegex, ProteinDbLoader.UniprotGeneNameRegex, ProteinDbLoader.UniprotOrganismRegex, out var un2).Select(p => p.NonVariantProtein).Distinct();
var newProts = ProteinAnnotation.CombineAndAnnotateProteins(uniprot, nonVariantProts.ToList());
ProteinDbWriter.WriteXmlDatabase(null, newProts, outxml);
string outfasta = Path.Combine(Path.GetDirectoryName(destinationXmlPath), Path.GetFileNameWithoutExtension(destinationXmlPath) + ".fasta");
string outfastaWithDecoys = Path.Combine(Path.GetDirectoryName(destinationXmlPath), Path.GetFileNameWithoutExtension(destinationXmlPath) + ".withdecoys.fasta");
var prot = newProts.FirstOrDefault(p => p.Accession.Contains("_"));
var protsForFasta = newProts.SelectMany(p => p.GetVariantProteins()).Where(p => !p.BaseSequence.EndsWith('?')).ToList();
var decoyProtsForFasta = ProteinDbLoader.LoadProteinXML(destinationXmlPath, true, DecoyType.Reverse, uniprotPtms, false, null, out un).Where(p => !p.BaseSequence.EndsWith('?')).ToList();
ProteinDbWriter.WriteFastaDatabase(protsForFasta, outfasta, "|");
ProteinDbWriter.WriteFastaDatabase(decoyProtsForFasta, outfastaWithDecoys, "|");
File.WriteAllLines(outfastaWithDecoys, File.ReadAllLines(outfastaWithDecoys).Select(line => line.Replace("mz|DECOY_", "rev_mz|")));
return(outxml);
}
