/// <summary>
/// Fill reference allele from genome fasta file.
/// </summary>
/// <param name="snpItems"></param>
/// <param name="fastaFile"></param>
/// <param name="progress"></param>
public static void FillReferenceAlleleFromFasta(this IEnumerable <SNPItem> snpItems, string fastaFile, IProgressCallback progress = null)
{
if (progress == null)
{
progress = new ConsoleProgressCallback();
}
var dic = snpItems.ToGroupDictionary(m => m.Chrom);
progress.SetMessage("Filling reference allele from {0} file ...", fastaFile);
using (var sw = new StreamReader(fastaFile))
{
var ff = new FastaFormat();
Sequence seq;
while ((seq = ff.ReadSequence(sw)) != null)
{
progress.SetMessage("chromosome " + seq.Name + " ...");
var chr = HumanChromosomeToInt(seq.Name);
if (dic.ContainsKey(chr))
{
var snps = dic[chr];
foreach (var snp in snps)
{
snp.RefChar = char.ToUpper(seq.SeqString[snp.Position - 1]);
}
}
}
}
progress.SetMessage("Filling reference allele finished.");
}
public override IEnumerable <string> Process(string fileName)
{
string result = FileUtils.ChangeExtension(fileName, "") + "_" + name + new FileInfo(fileName).Extension;
FastaFormat format = new FastaFormat();
Progress.SetMessage("Processing " + fileName);
using (StreamReader sr = new StreamReader(fileName))
{
Progress.SetRange(0, sr.BaseStream.Length);
using (StreamWriter sw = new StreamWriter(result))
{
Sequence seq;
while ((seq = format.ReadSequence(sr)) != null)
{
if (Progress.IsCancellationPending())
{
throw new UserTerminatedException();
}
Progress.SetPosition(sr.BaseStream.Position);
if (nameRegex.Match(seq.Name).Success)
{
format.WriteSequence(sw, seq);
}
}
}
}
return(new string[] { result });
}
public static List <CoverageRegion> GetTargetCoverageRegion(ITargetBuilderOptions options, IProgressCallback progress, bool removeRegionWithoutSequence = true)
{
List <CoverageRegion> result;
if (options.TargetFile.EndsWith(".xml"))
{
result = GetTargetCoverageRegionFromXml(options, progress);
}
else
{
result = GetTargetCoverageRegionFromBed(options, progress);
}
var dic = result.ToGroupDictionary(m => m.Seqname);
progress.SetMessage("Filling sequence from {0}...", options.GenomeFastaFile);
using (var sr = new StreamReader(options.GenomeFastaFile))
{
var ff = new FastaFormat();
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
progress.SetMessage("Processing chromosome {0} ...", seq.Reference);
var seqname = seq.Name.StringAfter("chr");
List <CoverageRegion> lst;
if (dic.TryGetValue(seqname, out lst))
{
foreach (var l in lst)
{
l.Sequence = seq.SeqString.Substring((int)(l.Start - 1), (int)l.Length);
if (l.Strand == '+')
{
l.ReverseComplementedSequence = SequenceUtils.GetReverseComplementedSequence(l.Sequence);
}
}
}
}
}
if (removeRegionWithoutSequence)
{
result.RemoveAll(l => string.IsNullOrEmpty(l.Sequence));
}
progress.SetMessage("Filling sequence finished.");
var namemap = new MapReader(1, 12).ReadFromFile(options.RefgeneFile);
result.ForEach(m =>
{
var gene = m.Name.StringBefore("_utr3");
m.GeneSymbol = namemap.ContainsKey(gene) ? namemap[gene] : string.Empty;
});
return(result);
}
public static HashSet <string> GetContaminationAccessNumbers(IStringParser <string> acParser, string fastaFilename, string contaminationDescriptionPattern,
IProgressCallback progress)
{
HashSet <string> result = new HashSet <string>();
if (progress == null)
{
progress = new EmptyProgressCallback();
}
Regex reg = new Regex(contaminationDescriptionPattern, RegexOptions.IgnoreCase);
progress.SetMessage("Get contamination map from database ...");
var ff = new FastaFormat();
using (var sr = new StreamReader(fastaFilename))
{
progress.SetRange(1, sr.BaseStream.Length);
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
if (progress.IsCancellationPending())
{
throw new UserTerminatedException();
}
progress.SetPosition(sr.GetCharpos());
string ac = acParser.GetValue(seq.Name);
if (reg.Match(seq.Reference).Success)
{
result.Add(ac);
}
}
}
progress.SetMessage("Get contamination map from database finished.");
return(result);
}
private void ProcessFile(ref int index, StreamWriter sw, string fastaFile, bool isContaminant)
{
FastaFormat ff = new FastaFormat();
using (StreamReader sr = new StreamReader(fastaFile))
{
Progress.SetRange(0, sr.BaseStream.Length);
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
Progress.SetPosition(sr.BaseStream.Position);
if (isContaminant)
{
if (!seq.Reference.StartsWith("CON_"))
{
seq.Reference = "CON_" + seq.Reference;
}
}
if (!options.ReversedOnly)
{
ff.WriteSequence(sw, seq);
}
if (options.IsPseudoAminoacid)
{
options.PseudoAminoacidBuilder.Build(seq);
}
index++;
Sequence reversedSeq = GetReversedSequence(index, seq);
ff.WriteSequence(sw, reversedSeq);
}
}
}
public override IEnumerable <string> Process(string fileName)
{
FastaFormat ff = new FastaFormat();
var result = Path.ChangeExtension(fileName, ".dM.fasta");
using (StreamReader sr = new StreamReader(fileName))
using (StreamWriter sw = new StreamWriter(result))
{
Sequence seq;
Progress.SetRange(1, sr.BaseStream.Length);
while ((seq = ff.ReadSequence(sr)) != null)
{
Progress.SetPosition(StreamUtils.GetCharpos(sr));
if (seq.SeqString.StartsWith("M"))
{
seq.SeqString = seq.SeqString.Substring(1);
seq.Reference = seq.Name + " N-terminal-M-Removed " + seq.Description;
}
ff.WriteSequence(sw, seq);
}
}
return(new string[] { result });
}
public override IEnumerable <string> Process()
{
var paramFile = options.OutputFile + ".param";
options.SaveToFile(options.OutputFile + ".param");
var bedfile = new BedItemFile <BedItem>(6);
Progress.SetMessage("building chromosome name map ...");
var mitoName = "M";
Dictionary <string, string> chrNameMap = new Dictionary <string, string>();
var ff = new FastaFormat(int.MaxValue);
var faiFile = options.FastaFile + ".fai";
if (File.Exists(faiFile))
{
using (StreamReader sr = new StreamReader(faiFile))
{
string line;
while ((line = sr.ReadLine()) != null)
{
var name = line.Split('\t')[0];
chrNameMap[name] = name;
if (name.StartsWith("chr"))
{
chrNameMap[name.StringAfter("chr")] = name;
}
if (!name.StartsWith("chr"))
{
chrNameMap["chr" + name] = name;
}
if (name.Equals("chrMT") || name.Equals("MT"))
{
mitoName = "MT";
}
if (name.Equals("chrM") || name.Equals("M"))
{
mitoName = "M";
}
}
}
}
else
{
using (StreamReader sr = new StreamReader(options.FastaFile))
{
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
var name = seq.Name;
chrNameMap[name] = name;
if (name.StartsWith("chr"))
{
chrNameMap[name.StringAfter("chr")] = name;
}
if (!name.StartsWith("chr"))
{
chrNameMap["chr" + name] = name;
}
if (name.Equals("chrMT") || name.Equals("MT"))
{
mitoName = "MT";
}
if (name.Equals("chrM") || name.Equals("M"))
{
mitoName = "M";
}
}
}
}
var longMitoName = chrNameMap[mitoName];
Progress.SetMessage("mitochondral chromosome name = {0}", longMitoName);
var mirnas = new List <BedItem>();
if (File.Exists(options.MiRBaseFile))
{
Progress.SetMessage("Processing {0} ...", options.MiRBaseFile);
if (options.MiRBaseFile.EndsWith(".bed"))
{
mirnas = bedfile.ReadFromFile(options.MiRBaseFile);
mirnas.ForEach(m =>
{
m.Seqname = m.Seqname.StringAfter("chr");
m.Name = options.MiRBaseKey + ":" + m.Name;
});
}
else
{
using (var gf = new GtfItemFile(options.MiRBaseFile))
{
GtfItem item;
while ((item = gf.Next(options.MiRBaseKey)) != null)
{
BedItem loc = new BedItem();
loc.Seqname = item.Seqname.StringAfter("chr");
loc.Start = item.Start - 1;
loc.End = item.End;
loc.Name = options.MiRBaseKey + ":" + item.Attributes.StringAfter("Name=").StringBefore(";");
loc.Score = 1000;
loc.Strand = item.Strand;
mirnas.Add(loc);
}
}
}
Progress.SetMessage("{0} miRNA readed.", mirnas.Count);
}
List <BedItem> trnas = new List <BedItem>();
if (File.Exists(options.UcscTrnaFile))
{
//reading tRNA from ucsc table without mitocondrom tRNA
Progress.SetMessage("Processing {0} ...", options.UcscTrnaFile);
trnas = bedfile.ReadFromFile(options.UcscTrnaFile);
trnas.ForEach(m => m.Seqname = m.Seqname.StringAfter("chr"));
var removed = trnas.Where(m => (m.Seqname.Length > 1) && !m.Seqname.All(n => char.IsDigit(n))).ToList();
if (removed.Count != trnas.Count)
{
//remove the tRNA not from 1-22, X and Y
trnas.RemoveAll(m => (m.Seqname.Length > 1) && !m.Seqname.All(n => char.IsDigit(n)));
//mitocondrom tRNA will be extracted from ensembl gtf file
trnas.RemoveAll(m => m.Seqname.Equals("M") || m.Seqname.Equals("MT"));
}
trnas.ForEach(m => m.Name = GetTRNAName(m.Name));
Progress.SetMessage("{0} tRNA from ucsc readed.", trnas.Count);
if (File.Exists(options.UcscMatureTrnaFastaFile))
{
var seqs = SequenceUtils.Read(options.UcscMatureTrnaFastaFile);
foreach (var seq in seqs)
{
var tRNAName = GetTRNAName(seq.Name);
trnas.Add(new BedItem()
{
Seqname = seq.Name,
Start = 0,
End = seq.SeqString.Length,
Strand = '+',
Name = tRNAName,
Sequence = seq.SeqString
});
}
}
}
var others = new List <BedItem>();
if (File.Exists(options.EnsemblGtfFile))
{
//reading smallRNA/tRNA from ensembl gtf file
Progress.SetMessage("Processing {0} ...", options.EnsemblGtfFile);
using (var gf = new GtfItemFile(options.EnsemblGtfFile))
{
var biotypes = new HashSet <string>(SmallRNAConsts.Biotypes);
biotypes.Remove(SmallRNAConsts.miRNA);
GtfItem item;
int count = 0;
while ((item = gf.Next("gene")) != null)
{
string biotype;
if (item.Attributes.Contains("gene_biotype"))
{
biotype = item.Attributes.StringAfter("gene_biotype \"").StringBefore("\"");
}
else if (item.Attributes.Contains("gene_type"))
{
biotype = item.Attributes.StringAfter("gene_type \"").StringBefore("\"");
}
else
{
continue;
}
if (File.Exists(options.UcscTrnaFile) && biotype.Equals(SmallRNAConsts.tRNA))
{
continue;
}
if (biotype.Equals("Mt_tRNA"))
{
count++;
var gene_name = item.Attributes.Contains("gene_name") ? item.Attributes.StringAfter("gene_name \"").StringBefore("\"") : item.GeneId;
BedItem loc = new BedItem();
loc.Seqname = mitoName;
loc.Start = item.Start - 1;
loc.End = item.End;
loc.Name = string.Format(SmallRNAConsts.mt_tRNA + ":" + longMitoName + ".tRNA{0}-{1}", count, gene_name.StringAfter("-"));
loc.Score = 1000;
loc.Strand = item.Strand;
trnas.Add(loc);
}
else if (biotypes.Contains(biotype))
{
string seqName;
if (item.Seqname.ToLower().StartsWith("chr"))
{
seqName = item.Seqname.Substring(3);
}
else
{
seqName = item.Seqname;
}
if (seqName.Equals("M") || seqName.Equals("MT"))
{
seqName = mitoName;
}
//ignore all smallRNA coordinates on scaffold or contig.
//if (seqName.Length > 5)
//{
// continue;
//}
var gene_name = item.Attributes.StringAfter("gene_name \"").StringBefore("\"");
var lowGeneName = gene_name.ToLower();
if (lowGeneName.StartsWith("rny") || lowGeneName.Equals("y_rna"))
{
biotype = "yRNA";
}
BedItem loc = new BedItem();
loc.Seqname = seqName;
loc.Start = item.Start - 1;
loc.End = item.End;
//if (lowGeneName.EndsWith("_rrna") && loc.Length < 200)
//{
// biotype = "rRNA";
//}
loc.Name = biotype + ":" + gene_name + ":" + item.GeneId;
loc.Score = 1000;
loc.Strand = item.Strand;
others.Add(loc);
}
}
}
}
var all = new List <BedItem>();
all.AddRange(mirnas);
all.AddRange(trnas);
all.AddRange(others);
foreach (var bi in all)
{
if (chrNameMap.ContainsKey(bi.Seqname))
{
bi.Seqname = chrNameMap[bi.Seqname];
}
}
if (File.Exists(options.RRNAFile))
{
var seqs = SequenceUtils.Read(options.RRNAFile);
foreach (var seq in seqs)
{
all.Add(new BedItem()
{
Seqname = seq.Name,
Start = 0,
End = seq.SeqString.Length,
Strand = '+',
Name = "rRNA:" + SmallRNAConsts.rRNADB_KEY + seq.Name
});
}
}
Progress.SetMessage("Saving smallRNA coordinates to " + options.OutputFile + "...");
using (var sw = new StreamWriter(options.OutputFile))
{
foreach (var pir in SmallRNAConsts.Biotypes)
{
var locs = all.Where(m => m.Name.StartsWith(pir)).ToList();
Progress.SetMessage("{0} : {1}", pir, locs.Count);
GenomeUtils.SortChromosome(locs, m => m.Seqname, m => (int)m.Start);
foreach (var loc in locs)
{
sw.WriteLine(bedfile.GetValue(loc));
}
}
}
var miRNA_bed = FileUtils.ChangeExtension(options.OutputFile, ".miRNA.bed");
Progress.SetMessage("Saving miRNA coordinates to " + miRNA_bed + "...");
using (var sw = new StreamWriter(miRNA_bed))
{
var pir = SmallRNAConsts.miRNA;
var locs = all.Where(m => m.Name.StartsWith(pir)).ToList();
Progress.SetMessage("{0} : {1}", pir, locs.Count);
GenomeUtils.SortChromosome(locs, m => m.Seqname, m => (int)m.Start);
foreach (var loc in locs)
{
sw.WriteLine(bedfile.GetValue(loc));
}
}
Progress.SetMessage("Saving smallRNA miss1 coordinates to " + options.OutputFile + ".miss1 ...");
using (var sw = new StreamWriter(options.OutputFile + ".miss1"))
{
foreach (var pir in SmallRNAConsts.Biotypes)
{
if (pir == SmallRNABiotype.lincRNA.ToString() || pir == SmallRNABiotype.lncRNA.ToString())
{
continue;
}
var locs = all.Where(m => m.Name.StartsWith(pir)).ToList();
locs.RemoveAll(l => l.Name.Contains(SmallRNAConsts.rRNADB_KEY));
Progress.SetMessage("{0} : {1}", pir, locs.Count);
GenomeUtils.SortChromosome(locs, m => m.Seqname, m => (int)m.Start);
foreach (var loc in locs)
{
sw.WriteLine(bedfile.GetValue(loc));
}
}
}
Progress.SetMessage("Saving smallRNA miss1 coordinates to " + options.OutputFile + ".miss0 ...");
using (var sw = new StreamWriter(options.OutputFile + ".miss0"))
{
foreach (var pir in SmallRNAConsts.Biotypes)
{
if (pir != SmallRNABiotype.lincRNA.ToString() && pir != SmallRNABiotype.lncRNA.ToString() && pir != SmallRNABiotype.rRNA.ToString())
{
continue;
}
var locs = all.Where(m => m.Name.StartsWith(pir)).ToList();
if (pir == SmallRNABiotype.rRNA.ToString())
{
locs.RemoveAll(l => !l.Name.Contains(SmallRNAConsts.rRNADB_KEY));
}
Progress.SetMessage("{0} : {1}", pir, locs.Count);
GenomeUtils.SortChromosome(locs, m => m.Seqname, m => (int)m.Start);
foreach (var loc in locs)
{
sw.WriteLine(bedfile.GetValue(loc));
}
}
}
var summaryFile = options.OutputFile + ".info";
Progress.SetMessage("Writing summary to " + summaryFile + "...");
using (var sw = new StreamWriter(summaryFile))
{
sw.WriteLine("Biotype\tCount");
all.ConvertAll(m => m.Name).Distinct().GroupBy(m => m.StringBefore(":")).OrderByDescending(m => m.Count()).ToList().ForEach(m => sw.WriteLine("{0}\t{1}", m.Key, m.Count()));
}
var result = new List <string>(new[] { options.OutputFile });
var fasta = Path.ChangeExtension(options.OutputFile, ".fasta");
if ((File.Exists(options.UcscTrnaFile) && File.Exists(options.UcscMatureTrnaFastaFile)) || File.Exists(options.RRNAFile))
{
result.Add(fasta);
using (var sw = new StreamWriter(fasta))
{
string line;
using (var sr = new StreamReader(options.FastaFile))
{
while ((line = sr.ReadLine()) != null)
{
sw.WriteLine(line);
}
}
if (File.Exists(options.UcscTrnaFile) && File.Exists(options.UcscMatureTrnaFastaFile))
{
using (var sr = new StreamReader(options.UcscMatureTrnaFastaFile))
{
while ((line = sr.ReadLine()) != null)
{
sw.WriteLine(line);
}
}
}
if (File.Exists(options.RRNAFile))
{
using (var sr = new StreamReader(options.RRNAFile))
{
while ((line = sr.ReadLine()) != null)
{
sw.WriteLine(line);
}
}
}
}
}
var faFile = options.OutputFile + ".fa";
Progress.SetMessage("Extracting sequence from " + options.FastaFile + "...");
var b2foptions = new Bed2FastaProcessorOptions()
{
GenomeFastaFile = options.FastaFile,
InputFile = options.OutputFile,
OutputFile = faFile,
KeepChrInName = false,
};
if (!File.Exists(options.UcscMatureTrnaFastaFile))
{
b2foptions.AcceptName = m => m.StartsWith(SmallRNAConsts.miRNA) || m.StartsWith(SmallRNAConsts.mt_tRNA) || m.StartsWith(SmallRNAConsts.tRNA);
}
else
{
b2foptions.AcceptName = m => m.StartsWith(SmallRNAConsts.miRNA) || m.StartsWith(SmallRNAConsts.mt_tRNA);
}
new Bed2FastaProcessor(b2foptions)
{
Progress = this.Progress
}.Process();
if (File.Exists(options.UcscMatureTrnaFastaFile))
{
Progress.SetMessage("Extracting sequence from " + options.UcscMatureTrnaFastaFile + " ...");
using (var sw = new StreamWriter(faFile, true))
{
foreach (var tRNA in trnas)
{
if (!string.IsNullOrEmpty(tRNA.Sequence))
{
sw.WriteLine(">{0}", tRNA.Name);
sw.WriteLine("{0}", tRNA.Sequence);
}
}
}
}
return(result);
}
public override IEnumerable <string> Process(string filename)
{
FastaFormat ff = new FastaFormat();
Digest digest = new Digest();
digest.DigestProtease = ProteaseManager.FindOrCreateProtease("Trypsin", true, "RK", "P");
digest.MaxMissedCleavages = 1;
NGlycanFilter filter = new NGlycanFilter();
digest.Filter = filter;
string resultFile = filename + ".nglycan";
Dictionary <string, NGlycanValue> peptideProteinMap = new Dictionary <string, NGlycanValue>();
using (StreamReader sr = new StreamReader(filename))
{
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
digest.ProteinSequence = seq;
digest.AddDigestFeatures();
if (seq.Annotation.ContainsKey(Digest.PEPTIDE_FEATURE_TYPE))
{
bool[] isGlycans = filter.IsNglycan;
List <DigestPeptideInfo> nglycanPeptides = (List <DigestPeptideInfo>)seq.Annotation[Digest.PEPTIDE_FEATURE_TYPE];
foreach (DigestPeptideInfo dpi in nglycanPeptides)
{
if (!peptideProteinMap.ContainsKey(dpi.PeptideSeq))
{
StringBuilder sb = new StringBuilder();
for (int i = 0; i < dpi.PeptideSeq.Length; i++)
{
if (isGlycans[dpi.PeptideLoc.Min - 1 + i])
{
sb.Append(1);
}
else
{
sb.Append(0);
}
}
NGlycanValue value = new NGlycanValue();
value.NGlycanSites = sb.ToString();
peptideProteinMap[dpi.PeptideSeq] = value;
}
peptideProteinMap[dpi.PeptideSeq].Proteins.Add(parser.GetValue(dpi.ProteinName));
}
}
}
}
List <string> peptides = new List <string>(peptideProteinMap.Keys);
peptides.Sort();
using (StreamWriter sw = new StreamWriter(resultFile))
{
foreach (string pep in peptides)
{
NGlycanValue value = peptideProteinMap[pep];
sw.Write(pep + "\t" + value.NGlycanSites + "\t");
bool bFirst = true;
foreach (string protein in value.Proteins)
{
if (bFirst)
{
bFirst = false;
sw.Write(protein);
}
else
{
sw.Write(" ! " + protein);
}
}
sw.WriteLine();
}
}
return(new[] { resultFile });
}
public static void FillSequenceFromFasta(IStringParser <string> acParser, string fastaFilename, IIdentifiedResult t,
IProgressCallback progress)
{
if (progress == null)
{
progress = new EmptyProgressCallback();
}
progress.SetMessage("Initializing accessNumber/protein map ...");
var acMap = new Dictionary <string, IIdentifiedProtein>();
foreach (IIdentifiedProteinGroup group in t)
{
foreach (IIdentifiedProtein protein in group)
{
string ac = acParser.GetValue(protein.Name);
if (acMap.ContainsKey(ac))
{
throw new Exception("Duplicate access number " + ac);
}
acMap[ac] = protein;
if (ac != protein.Name)
{
if (acMap.ContainsKey(protein.Name))
{
throw new Exception("Duplicate access number " + protein.Name);
}
acMap[protein.Name] = protein;
}
}
}
progress.SetMessage("Filling sequence from database ...");
var ff = new FastaFormat();
using (var sr = new StreamReader(fastaFilename))
{
progress.SetRange(1, sr.BaseStream.Length);
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
if (progress.IsCancellationPending())
{
throw new UserTerminatedException();
}
progress.SetPosition(sr.BaseStream.Position);
string ac = acParser.GetValue(seq.Name);
if (acMap.ContainsKey(ac))
{
IIdentifiedProtein protein = acMap[ac];
protein.Name = seq.Name.Replace("/", " ");
protein.Description = seq.Description.Replace("\t", " ").Replace("/", " ");
protein.Sequence = seq.SeqString;
}
}
}
var failed = acMap.Values.Where(l => l.Sequence == null).ToList();
if (failed.Count > 0)
{
var proteinNames = failed.ConvertAll(l => l.Name).ToArray();
if (!proteinNames.All(l => l.StartsWith("XXX_")))
{
throw new Exception(string.Format("Couldn't find sequence of following protein(s), change access number pattern or select another database\n{0}", proteinNames.Merge("/")));
}
}
progress.SetMessage("Fill sequence from database finished.");
}
public override IEnumerable <string> Process(string fileName)
{
var result = new List <string>();
string[] acLines = File.ReadAllLines(fileName);
var acs = new HashSet <string>();
foreach (var acline in acLines)
{
string ac;
if (!parser.TryParse(acline, out ac))
{
ac = acline;
}
acs.Add(ac);
}
var findAcs = new HashSet <string>();
var resultFile = fileName + ".fasta";
result.Add(resultFile);
var ff = new FastaFormat();
using (StreamWriter sw = new StreamWriter(resultFile))
using (StreamReader sr = new StreamReader(database))
{
Progress.SetRange(0, sr.BaseStream.Length);
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
Progress.SetPosition(sr.BaseStream.Position);
string curAc;
if (!parser.TryParse(seq.Name, out curAc))
{
curAc = seq.Name;
}
if (acs.Contains(curAc))
{
findAcs.Add(curAc);
if (this.replaceName)
{
seq.Reference = curAc;
}
ff.WriteSequence(sw, seq);
}
}
}
acs.ExceptWith(findAcs);
var missFile = fileName + ".miss";
if (acs.Count > 0)
{
using (StreamWriter sw = new StreamWriter(missFile))
{
foreach (var ac in acs)
{
sw.WriteLine(ac);
}
}
result.Add(missFile);
}
else if (File.Exists(missFile))
{
File.Delete(missFile);
}
return(result);
}
/// <summary>
/// 读取fasta文件，进行数据处理。
/// </summary>
/// <param name="fileName"></param>
/// <returns></returns>
public override IEnumerable <string> Process(string fastaFile)
{
HashSet <string> result = new HashSet <string>();
var ff = new FastaFormat();
using (StreamReader sr = new StreamReader(fastaFile))
{
Progress.SetRange(0, sr.BaseStream.Length);
var aas = new Aminoacids();
Predicate <string> aaFilter = m =>
{
foreach (var aa in ignoreAminoacids)
{
if (m.Contains(aa))
{
return(false);
}
}
return(true);
};
Predicate <string> lengthFilter = m => m.Length >= minLength;
Predicate <string> massFilter = m =>
{
var mass = aas.MonoPeptideMass(m);
return(mass >= minMass && mass <= maxMass);
};
Predicate <string> filter = m => aaFilter(m) && lengthFilter(m) && massFilter(m);
List <Digest> digs = new List <Digest>();
foreach (var protease in proteases)
{
var dig = new Digest();
dig.DigestProtease = protease;
dig.MaxMissedCleavages = maxMissCleavage;
digs.Add(dig);
}
Sequence seq;
Progress.SetMessage("Digesting sequences ...");
while ((seq = ff.ReadSequence(sr)) != null)
{
Progress.SetPosition(sr.GetCharpos());
if (Progress.IsCancellationPending())
{
throw new UserTerminatedException();
}
HashSet <string> curseqs = new HashSet <string>();
curseqs.Add(seq.SeqString);
foreach (var dig in digs)
{
var last = curseqs;
curseqs = new HashSet <string>();
foreach (var curseq in last)
{
var pro = new Sequence(curseq, curseq);
dig.ProteinSequence = pro;
dig.AddDigestFeatures();
var infos = pro.GetDigestPeptideInfo();
infos.ForEach(m =>
{
if (filter(m.PeptideSeq))
{
curseqs.Add(m.PeptideSeq);
}
});
}
}
result.UnionWith(curseqs);
}
}
Progress.SetMessage("Sorting sequences ...");
var peps = new List <string>(result);
peps.Sort((m1, m2) =>
{
var res = m1.Length.CompareTo(m2.Length);
if (res == 0)
{
res = m1.CompareTo(m2);
}
return(res);
});
var resultFile = fastaFile + ".pep";
using (StreamWriter sw = new StreamWriter(resultFile))
{
peps.ForEach(m => sw.WriteLine(m));
}
return(new[] { resultFile });
}
public override IEnumerable <string> Process()
{
var srItems = SequenceRegionUtils.GetSequenceRegions(options.InputFile).Where(m => options.AcceptName(m.Name)).ToList();
srItems = (from sr in srItems.GroupBy(m => m.Name)
select sr.First()).ToList();
var keepChrInName = options.KeepChrInName && srItems.Any(m => m.Name.StartsWith("chr"));
if (!keepChrInName)
{
srItems.ForEach(m => m.Seqname = m.Seqname.StringAfter("chr"));
}
var srMap = srItems.ToGroupDictionary(m => m.Seqname);
var ff = new FastaFormat(int.MaxValue);
using (StreamWriter sw = new StreamWriter(options.OutputFile))
{
using (StreamReader sr = new StreamReader(options.GenomeFastaFile))
{
Sequence seq;
while ((seq = ff.ReadSequence(sr)) != null)
{
Progress.SetMessage("processing " + seq.Name + " ...");
var name = seq.Name;
if (!keepChrInName)
{
name = name.StringAfter("chr");
}
List <GtfItem> items;
if (!srMap.TryGetValue(name, out items))
{
if (name.Equals("M"))
{
name = "MT";
srMap.TryGetValue(name, out items);
}
else if (name.Equals("chrM"))
{
name = "chrMT";
srMap.TryGetValue(name, out items);
}
else if (name.Equals("MT"))
{
name = "M";
srMap.TryGetValue(name, out items);
}
else if (name.Equals("chrMT"))
{
name = "chrM";
srMap.TryGetValue(name, out items);
}
}
if (items != null)
{
Progress.SetMessage(" there are {0} entries in {1} ...", items.Count, name);
foreach (var item in items)
{
if (item.Start - 1 + item.Length >= seq.SeqString.Length)
{
throw new Exception(string.Format("{0} exceed chromosome {1} length {2}", item, name, seq.SeqString.Length));
}
var newseq = seq.SeqString.Substring((int)item.Start - 1, (int)item.Length);
if (item.Strand == '-')
{
newseq = SequenceUtils.GetReverseComplementedSequence(newseq);
}
newseq = newseq.ToUpper();
var newname = string.Format("{0} {1} {2}", item.Name, item.GetLocationWithoutStrand(), item.Strand);
var entry = new Sequence(newname, newseq);
ff.WriteSequence(sw, entry);
}
}
}
}
}
return(new string[] { options.OutputFile });
}