private void GenerateMetaFiles()
{
if (File.Exists(_seqFilePath))
{
File.Delete(_seqFilePath);
}
if (File.Exists(_annoFilePath))
{
File.Delete(_annoFilePath);
}
using (var seqWriter = new BinaryWriter(File.Open(_seqFilePath, FileMode.CreateNew)))
using (var annoWriter = new StreamWriter(_annoFilePath))
{
// Read
var reader = new FastaFileReader();
if (!reader.OpenFile(_databaseFilePath))
{
return;
}
long offset = 0;
while (reader.ReadNextProteinEntry())
{
var name = reader.ProteinName;
var description = reader.ProteinDescription;
var sequence = (char)Delimiter + reader.ProteinSequence;
var length = reader.ProteinSequence.Length;
seqWriter.Write(Encoding.GetBytes(sequence));
annoWriter.WriteLine("{0}{1}{2}{3}{4}{5}{6}",
offset,
AnnotationDelimiter,
length,
AnnotationDelimiter,
name.Replace(AnnotationDelimiter, '-'),
AnnotationDelimiter,
description);
offset += sequence.Length;
}
seqWriter.Write((char)Delimiter);
// write file format Id
seqWriter.Write(FileFormatId);
// writer last write time hash
var hashCode = File.GetLastWriteTimeUtc(_databaseFilePath).GetHashCode();
seqWriter.Write(hashCode);
annoWriter.Write("{0}{1}{2}", FileFormatId, AnnotationDelimiter, hashCode);
reader.CloseFile();
}
}
/// <summary>
/// Create the decoy version of this databse
/// </summary>
/// <param name="enzyme"></param>
/// <param name="shuffle"></param>
public void CreateDecoyDatabase(Enzyme enzyme, bool shuffle)
{
var reader = new FastaFileReader();
if (!reader.OpenFile(_databaseFilePath))
{
return;
}
var decoyDatabaseFileName = GetDecoyDatabasePath(enzyme, shuffle);
Console.WriteLine("Creating " + decoyDatabaseFileName);
using (var decoyWriter = new StreamWriter(decoyDatabaseFileName))
{
while (reader.ReadNextProteinEntry())
{
var name = reader.ProteinName;
var description = reader.ProteinDescription;
var sequence = reader.ProteinSequence;
decoyWriter.WriteLine(">{0}_{1} {2}", FastaDatabaseConstants.DecoyProteinPrefix, name, description);
if (!shuffle)
{
// Reversed protein sequence
var decoySequence = new StringBuilder();
for (var i = sequence.Length - 1; i >= 0; i--)
{
var residue = sequence[i];
if (enzyme != null && enzyme.Residues.Length > 0 && enzyme.IsCleavable(residue) && decoySequence.Length > 0)
{
var residueToBeReplaced = decoySequence[decoySequence.Length - 1];
decoySequence.Remove(decoySequence.Length - 1, 1);
decoySequence.Append((char)residue);
decoySequence.Append(residueToBeReplaced);
}
else
{
decoySequence.Append((char)residue);
}
}
decoyWriter.WriteLine(decoySequence);
}
else
{
// Shuffled protein sequences
decoyWriter.WriteLine(SimpleStringProcessing.Mutate(SimpleStringProcessing.Shuffle(sequence), NumMutations));
}
}
reader.CloseFile();
}
}
private void GenerateMetaFiles()
{
if (File.Exists(_seqFilePath))
{
File.Delete(_seqFilePath);
}
if (File.Exists(_annoFilePath))
{
File.Delete(_annoFilePath);
}
// Keys are protein name
// Values track the number of times the name has been encountered, the number of residues, and a SHA1 hash of the residues
var proteinNamesAndStats = new Dictionary <string, ProteinHashInfo>(StringComparer.InvariantCultureIgnoreCase);
using (var seqWriter = new BinaryWriter(File.Open(_seqFilePath, FileMode.CreateNew)))
using (var annoWriter = new StreamWriter(_annoFilePath))
{
// Read
var reader = new FastaFileReader();
if (!reader.OpenFile(_databaseFilePath))
{
return;
}
long offset = 0;
while (reader.ReadNextProteinEntry())
{
var name = reader.ProteinName;
var description = reader.ProteinDescription;
var sequence = (char)FastaDatabaseConstants.Delimiter + reader.ProteinSequence;
var length = sequence.Length;
var proteinInfoCurrent = new ProteinHashInfo(sequence);
if (proteinNamesAndStats.TryGetValue(name, out var proteinInfoCached))
{
// Duplicate name; either skip this protein or rename it
if (proteinInfoCached.SequenceLength == proteinInfoCurrent.SequenceLength &&
string.Equals(proteinInfoCached.SequenceHash, proteinInfoCurrent.SequenceHash))
{
// Duplicate protein; skip it
continue;
}
name += "_Duplicate" + proteinInfoCached.ObservationCount.ToString("00");
proteinInfoCached.ObservationCount++;
}
proteinNamesAndStats.Add(name, proteinInfoCurrent);
seqWriter.Write(Encoding.GetBytes((string)sequence));
annoWriter.WriteLine("{0}{1}{2}{3}{4}{5}{6}",
offset, FastaDatabaseConstants.AnnotationDelimiter,
length, FastaDatabaseConstants.AnnotationDelimiter,
name.Replace(FastaDatabaseConstants.AnnotationDelimiter, '-'), FastaDatabaseConstants.AnnotationDelimiter,
description);
offset += sequence.Length;
}
seqWriter.Write((char)FastaDatabaseConstants.Delimiter);
// write file format Id
seqWriter.Write(FileFormatId);
// writer last write time hash
var hashCode = File.GetLastWriteTimeUtc(_databaseFilePath).GetHashCode();
seqWriter.Write(hashCode);
annoWriter.Write("{0}{1}{2}", FileFormatId, FastaDatabaseConstants.AnnotationDelimiter, hashCode);
reader.CloseFile();
}
}
/// <summary>
/// Create the decoy version of this databse
/// </summary>
/// <param name="enzyme"></param>
/// <param name="shuffle"></param>
public void CreateDecoyDatabase(Enzyme enzyme, bool shuffle)
{
var reader = new FastaFileReader();
if (!reader.OpenFile(_databaseFilePath))
return;
var decoyDatabaseFileName = GetDecoyDatabasePath(enzyme, shuffle);
Console.WriteLine("Creating " + decoyDatabaseFileName);
using (var decoyWriter = new StreamWriter(decoyDatabaseFileName))
{
while (reader.ReadNextProteinEntry())
{
var name = reader.ProteinName;
var description = reader.ProteinDescription;
var sequence = reader.ProteinSequence;
decoyWriter.WriteLine(">{0}_{1} {2}", DecoyProteinPrefix, name, description);
if (!shuffle)
{
// Reversed protein sequence
var decoySequence = new StringBuilder();
for (var i = sequence.Length - 1; i >= 0; i--)
{
var residue = sequence[i];
if (enzyme != null && enzyme.Residues.Length > 0 && enzyme.IsCleavable(residue) && decoySequence.Length > 0)
{
var residueToBeReplaced = decoySequence[decoySequence.Length - 1];
decoySequence.Remove(decoySequence.Length - 1, 1);
decoySequence.Append(residue);
decoySequence.Append(residueToBeReplaced);
}
else
{
decoySequence.Append(residue);
}
}
decoyWriter.WriteLine(decoySequence);
}
else
{
// Shuffled protein sequences
decoyWriter.WriteLine(SimpleStringProcessing.Mutate(SimpleStringProcessing.Shuffle(sequence), NumMutations));
}
}
reader.CloseFile();
}
}
private void GenerateMetaFiles()
{
if (File.Exists(_seqFilePath))
File.Delete(_seqFilePath);
if (File.Exists(_annoFilePath))
File.Delete(_annoFilePath);
using (var seqWriter = new BinaryWriter(File.Open(_seqFilePath, FileMode.CreateNew)))
using (var annoWriter = new StreamWriter(_annoFilePath))
{
// Read
var reader = new FastaFileReader();
if (!reader.OpenFile(_databaseFilePath))
return;
long offset = 0;
while (reader.ReadNextProteinEntry())
{
var name = reader.ProteinName;
var description = reader.ProteinDescription;
var sequence = (char)Delimiter + reader.ProteinSequence;
var length = reader.ProteinSequence.Length;
seqWriter.Write(Encoding.GetBytes(sequence));
annoWriter.WriteLine("{0}{1}{2}{3}{4}{5}{6}",
offset,
AnnotationDelimiter,
length,
AnnotationDelimiter,
name.Replace(AnnotationDelimiter, '-'),
AnnotationDelimiter,
description);
offset += sequence.Length;
}
seqWriter.Write((char)Delimiter);
// write file format Id
seqWriter.Write(FileFormatId);
// writer last write time hash
var hashCode = File.GetLastWriteTimeUtc(_databaseFilePath).GetHashCode();
seqWriter.Write(hashCode);
annoWriter.Write("{0}{1}{2}", FileFormatId, AnnotationDelimiter, hashCode);
reader.CloseFile();
}
}