/// <summary>
/// Private Constructor for clone method.
/// </summary>
/// <param name="other">CrossReferenceLink instance to clone.</param>
private CrossReferenceLink(CrossReferenceLink other)
{
Type = other.Type;
Numbers = new List <string>(other.Numbers);
}
/// <summary>
/// Helper method to parse the metadata of gen bank data
/// </summary>
/// <param name="metadata">Metadata object</param>
/// <param name="cellRange">Range of cells</param>
/// <param name="rowIndex">Current index of row</param>
/// <returns>Index of row</returns>
private static int ParseGenBankMetadata(GenBankMetadata metadata, object[,] cellRange, int rowIndex)
{
string message = string.Empty;
string key;
string subKey;
string value = string.Empty;
while (rowIndex < cellRange.GetLength(0))
{
if (null != cellRange[rowIndex, KeyColumnIndex] && !string.IsNullOrWhiteSpace(cellRange[rowIndex, KeyColumnIndex].ToString()))
{
key = cellRange[rowIndex, KeyColumnIndex].ToString().ToUpperInvariant();
if (key.Equals(FEATURES))
{
break;
}
}
else
{
rowIndex++;
continue;
}
subKey = cellRange[rowIndex, SubKeyColumnIndex] != null ? cellRange[rowIndex, SubKeyColumnIndex].ToString() : string.Empty;
value = cellRange[rowIndex, ValueColumnIndex] != null ? cellRange[rowIndex, ValueColumnIndex].ToString() : string.Empty;
string[] tokens;
switch (key)
{
case LOCUS:
rowIndex = ParseLocus(metadata, cellRange, rowIndex);
rowIndex--;
break;
case DEFINITION:
metadata.Definition = value;
break;
case ACCESSION:
metadata.Accession = new GenBankAccession();
if (string.IsNullOrWhiteSpace(value))
{
message = String.Format(
CultureInfo.InvariantCulture,
Resources.UnrecognizedGenBankMetadataFormat,
ACCESSION);
throw new FormatException(message);
}
string[] accessions = value.Split(' ');
metadata.Accession.Primary = accessions[0];
for (int i = 1; i < accessions.Length; i++)
{
metadata.Accession.Secondary.Add(accessions[i]);
}
break;
case DBLINK:
if (!string.IsNullOrWhiteSpace(value))
{
break;
}
tokens = value.Split(':');
if (tokens.Length == 2)
{
if (metadata.DbLinks == null)
{ metadata.DbLinks = new List<CrossReferenceLink>(2); }
var curLink = new CrossReferenceLink();
if (string.Compare(tokens[0],
CrossReferenceType.Project.ToString(),
StringComparison.OrdinalIgnoreCase) == 0)
{
curLink.Type = CrossReferenceType.Project;
}
else if (string.Compare(tokens[0],
CrossReferenceType.BioProject.ToString(),
StringComparison.OrdinalIgnoreCase) == 0)
{
curLink.Type = CrossReferenceType.BioProject;
}
else
{
curLink.Type = CrossReferenceType.TraceAssemblyArchive;
}
tokens = tokens[1].Split(',');
for (int i = 0; i < tokens.Length; i++)
{
curLink.Numbers.Add(tokens[i]);
}
metadata.DbLinks.Add(curLink);
}
break;
case DBSOURCE:
metadata.DbSource = value;
break;
case VERSION:
if (string.IsNullOrWhiteSpace(value))
{
break;
}
tokens = value.Split(new char[] { ' ' },
StringSplitOptions.RemoveEmptyEntries);
// first token contains accession and version
Match m = Regex.Match(tokens[0], @"^(?<accession>\w+)\.(?<version>\d+)$");
metadata.Version = new GenBankVersion();
if (m.Success)
{
metadata.Version.Version = m.Groups["version"].Value;
// The first token in the data from the accession line is referred to as
// the primary accession number, and should be the one used here in the
// version line.
metadata.Version.Accession = m.Groups["accession"].Value;
}
if (tokens.Length > 1)
{
// second token contains primary ID
m = Regex.Match(tokens[1], @"^GI:(?<primaryID>.*)");
if (m.Success)
{
metadata.Version.GiNumber = m.Groups["primaryID"].Value;
}
}
break;
case SEGMENT:
if (string.IsNullOrWhiteSpace(value))
{
break;
}
tokens = value.Split(" of ".ToArray(), StringSplitOptions.RemoveEmptyEntries);
if (tokens.Length == 2)
{
int current;
int count;
if (int.TryParse(tokens[0], out current))
{
if (int.TryParse(tokens[1], out count))
{
metadata.Segment = new SequenceSegment();
metadata.Segment.Current = current;
metadata.Segment.Count = count;
}
}
}
if (metadata.Segment == null)
{
message = String.Format(
CultureInfo.InvariantCulture,
Resources.UnrecognizedGenBankMetadataFormat,
ACCESSION);
throw new FormatException(message);
}
break;
case KEYWORDS:
metadata.Keywords = value;
break;
case SOURCE:
rowIndex = ParseSource(metadata, cellRange, rowIndex);
rowIndex--;
break;
case REFERENCE:
rowIndex = ParseReference(metadata, cellRange, rowIndex);
rowIndex--;
break;
case PRIMARY:
metadata.Primary = value;
break;
case COMMENT:
if (!string.IsNullOrWhiteSpace(value))
{
tokens = value.Split(Environment.NewLine.ToCharArray(), StringSplitOptions.RemoveEmptyEntries);
foreach (string str in tokens)
{
metadata.Comments.Add(str);
}
}
break;
}
rowIndex++;
}
return rowIndex;
}
/// <summary>
/// Parses the GenBank headers from the GenBank file.
/// parses everything before the features section
/// </summary>
/// <param name="sequence">The sequence.</param>
/// <param name="noOfSequence">The current sequence index.</param>
/// <param name="line">parse line</param>
/// <param name="stream">The stream reader.</param>
/// <returns>The parsed line.</returns>
private string ParseHeaders(ref Sequence sequence, int noOfSequence, string line, StreamReader stream)
{
GenBankMetadata metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
string data;
string[] tokens;
// only allow one locus line
bool haveParsedLocus = false;
string lineData;
if (noOfSequence == 0)
{
line = string.Empty;
line = GoToNextLine(line, stream);
}
// parse until we hit the features or sequence section
bool haveFinishedHeaders = false;
while ((line != null) && !haveFinishedHeaders)
{
switch (GetLineHeader(line, DataIndent))
{
case "LOCUS":
if (haveParsedLocus)
{
string message = String.Format(CultureInfo.CurrentCulture, Properties.Resource.ParserSecondLocus);
Trace.Report(message);
throw new InvalidDataException(message);
}
line = ParseLocusByTokens(line, ref sequence, stream);
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
haveParsedLocus = true;
// don't go to next line; current line still needs to be processed
break;
case "VERSION":
lineData = GetLineData(line, DataIndent);
tokens = lineData.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
// first token contains accession and version
Match m = Regex.Match(tokens[0], @"^(?<accession>\w+)\.(?<version>\d+)$");
metadata.Version = new GenBankVersion();
if (m.Success)
{
metadata.Version.Version = m.Groups["version"].Value;
// The first token in the data from the accession line is referred to as
// the primary accession number, and should be the one used here in the
// version line.
string versionLineAccession = m.Groups["accession"].Value;
if (metadata.Accession == null)
{
ApplicationLog.WriteLine("WARN: VERSION processed before ACCESSION");
}
else
{
if (!versionLineAccession.Equals(metadata.Accession.Primary))
{
ApplicationLog.WriteLine("WARN: VERSION tag doesn't match ACCESSION");
}
else
{
metadata.Version.Accession = metadata.Accession.Primary;
}
}
}
if (tokens.Length > 1)
{
// second token contains primary ID
m = Regex.Match(tokens[1], @"^GI:(?<primaryID>.*)");
if (m.Success)
{
metadata.Version.GiNumber = m.Groups["primaryID"].Value;
}
}
line = GoToNextLine(line, stream);
break;
case "PROJECT":
lineData = GetLineData(line, DataIndent);
tokens = lineData.Split(':');
if (tokens.Length == 2)
{
metadata.Project = new ProjectIdentifier { Name = tokens[0] };
tokens = tokens[1].Split(',');
for (int i = 0; i < tokens.Length; i++)
{
metadata.Project.Numbers.Add(tokens[i]);
}
}
else
{
ApplicationLog.WriteLine("WARN: unexpected PROJECT header: " + line);
}
line = GoToNextLine(line, stream);
break;
case "SOURCE":
line = ParseSource(line, ref sequence, stream);
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
// don't go to next line; current line still needs to be processed
break;
case "REFERENCE":
line = ParseReferences(line, ref sequence, stream); // can encounter more than one
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
// don't go to next line; current line still needs to be processed
break;
case "COMMENT":
line = ParseComments(line, ref sequence, stream); // can encounter more than one
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
// don't go to next line; current line still needs to be processed
break;
case "PRIMARY":
// This header is followed by sequence info in a table format that could be
// stored in a custom object. The first line contains column headers.
// For now, just validate the presence of the headers, and save the data
// as a string.
lineData = GetLineData(line, DataIndent);
tokens = lineData.Split("\t ".ToCharArray(), StringSplitOptions.RemoveEmptyEntries);
// Validating for minimum two headers.
if (tokens.Length != 4)
{
string message = String.Format(
CultureInfo.CurrentCulture,
Properties.Resource.ParserPrimaryLineError,
line);
Trace.Report(message);
throw new InvalidDataException(message);
}
string primaryData = ParseMultiLineData(ref line, Environment.NewLine, DataIndent, stream);
metadata.Primary = primaryData;
// don't go to next line; current line still needs to be processed
break;
// all the following are extracted the same way - possibly multiline
case "DEFINITION":
metadata.Definition = ParseMultiLineData(ref line, " ", DataIndent, stream);
break;
case "ACCESSION":
data = ParseMultiLineData(ref line, " ", DataIndent, stream);
metadata.Accession = new GenBankAccession();
string[] accessions = data.Split(' ');
metadata.Accession.Primary = accessions[0];
for (int i = 1; i < accessions.Length; i++)
{
metadata.Accession.Secondary.Add(accessions[i]);
}
break;
case "DBLINK":
data = ParseMultiLineData(ref line, "\n", DataIndent, stream);
metadata.DbLinks = new List<CrossReferenceLink>();
foreach (string link in data.Split('\n'))
{
tokens = link.Split(':');
if (tokens.Length == 2)
{
CrossReferenceLink newLink = new CrossReferenceLink();
if (string.Compare(tokens[0], CrossReferenceType.Project.ToString(), StringComparison.OrdinalIgnoreCase) == 0)
{
newLink.Type = CrossReferenceType.Project;
}
else if (string.Compare(tokens[0], CrossReferenceType.BioProject.ToString(), StringComparison.OrdinalIgnoreCase) == 0)
{
newLink.Type = CrossReferenceType.BioProject;
}
else
{
newLink.Type = CrossReferenceType.None;
if (string.Compare(tokens[0], TraceAssemblyArchive, StringComparison.OrdinalIgnoreCase) == 0)
{
newLink.Type = CrossReferenceType.TraceAssemblyArchive;
}
}
tokens = tokens[1].Split(',');
for (int i = 0; i < tokens.Length; i++)
{
newLink.Numbers.Add(tokens[i]);
}
metadata.DbLinks.Add(newLink);
}
else
{
ApplicationLog.WriteLine("WARN: unexpected DBLINK header: " + line);
}
}
break;
case "DBSOURCE":
metadata.DbSource = ParseMultiLineData(ref line, " ", DataIndent, stream);
break;
case "KEYWORDS":
metadata.Keywords = ParseMultiLineData(ref line, " ", DataIndent, stream);
break;
case "SEGMENT":
data = ParseMultiLineData(ref line, " ", DataIndent, stream);
const string delimeter = "of";
tokens = data.Split(delimeter.ToCharArray(), StringSplitOptions.RemoveEmptyEntries);
if (tokens.Length == 2)
{
metadata.Segment = new SequenceSegment();
int outvalue;
if (int.TryParse(tokens[0].Trim(), out outvalue))
{
metadata.Segment.Current = outvalue;
}
else
{
ApplicationLog.WriteLine("WARN: unexpected SEGMENT header: " + line);
}
if (int.TryParse(tokens[1].Trim(), out outvalue))
{
metadata.Segment.Count = outvalue;
}
else
{
ApplicationLog.WriteLine("WARN: unexpected SEGMENT header: " + line);
}
}
else
{
ApplicationLog.WriteLine("WARN: unexpected SEGMENT header: " + line);
}
break;
// all the following indicate sections beyond the headers parsed by this method
case "FEATURES":
case "BASE COUNT":
case "ORIGIN":
case "CONTIG":
haveFinishedHeaders = true;
break;
default:
string lineHeader = GetLineHeader(line, DataIndent);
lineData = GetLineData(line, DataIndent);
ApplicationLog.WriteLine(ToString() + "WARN: unknown {0} -> {1}", lineHeader, lineData);
string errMessage = String.Format(
CultureInfo.CurrentCulture,
Properties.Resource.ParseHeaderError,
lineHeader);
Trace.Report(errMessage);
throw new InvalidDataException(errMessage);
}
}
// check for required features
if (!haveParsedLocus)
{
string message = string.Format(CultureInfo.CurrentCulture, Properties.Resource.INVALID_INPUT_FILE, Name);
Trace.Report(message);
throw new InvalidDataException(message);
}
return line;
}
/// <summary>
/// Parses the GenBank headers from the GenBank file.
/// parses everything before the features section
/// </summary>
/// <param name="sequence">The sequence.</param>
/// <param name="noOfSequence">The current sequence index.</param>
/// <param name="line">parse line</param>
/// <param name="stream">The stream reader.</param>
/// <returns>The parsed line.</returns>
private string ParseHeaders(ref Sequence sequence, int noOfSequence, string line, StreamReader stream)
{
GenBankMetadata metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
string data;
string[] tokens;
// only allow one locus line
bool haveParsedLocus = false;
string lineData;
if (noOfSequence == 0)
{
line = string.Empty;
line = GoToNextLine(line, stream);
}
// parse until we hit the features or sequence section
bool haveFinishedHeaders = false;
while ((line != null) && !haveFinishedHeaders)
{
switch (GetLineHeader(line, DataIndent))
{
case "LOCUS":
if (haveParsedLocus)
{
string message = String.Format(CultureInfo.CurrentCulture, Properties.Resource.ParserSecondLocus);
Trace.Report(message);
throw new InvalidDataException(message);
}
line = ParseLocusByTokens(line, ref sequence, stream);
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
haveParsedLocus = true;
// don't go to next line; current line still needs to be processed
break;
case "VERSION":
lineData = GetLineData(line, DataIndent);
tokens = lineData.Split(new[] { ' ' }, StringSplitOptions.RemoveEmptyEntries);
// first token contains accession and version
Match m = Regex.Match(tokens[0], @"^(?<accession>\w+)\.(?<version>\d+)$");
metadata.Version = new GenBankVersion();
if (m.Success)
{
metadata.Version.Version = m.Groups["version"].Value;
// The first token in the data from the accession line is referred to as
// the primary accession number, and should be the one used here in the
// version line.
string versionLineAccession = m.Groups["accession"].Value;
if (metadata.Accession == null)
{
ApplicationLog.WriteLine("WARN: VERSION processed before ACCESSION");
}
else
{
if (!versionLineAccession.Equals(metadata.Accession.Primary))
{
ApplicationLog.WriteLine("WARN: VERSION tag doesn't match ACCESSION");
}
else
{
metadata.Version.Accession = metadata.Accession.Primary;
}
}
}
if (tokens.Length > 1)
{
// second token contains primary ID
m = Regex.Match(tokens[1], @"^GI:(?<primaryID>.*)");
if (m.Success)
{
metadata.Version.GiNumber = m.Groups["primaryID"].Value;
}
}
line = GoToNextLine(line, stream);
break;
case "PROJECT":
lineData = GetLineData(line, DataIndent);
tokens = lineData.Split(':');
if (tokens.Length == 2)
{
metadata.Project = new ProjectIdentifier {
Name = tokens[0]
};
tokens = tokens[1].Split(',');
for (int i = 0; i < tokens.Length; i++)
{
metadata.Project.Numbers.Add(tokens[i]);
}
}
else
{
ApplicationLog.WriteLine("WARN: unexpected PROJECT header: " + line);
}
line = GoToNextLine(line, stream);
break;
case "SOURCE":
line = ParseSource(line, ref sequence, stream);
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
// don't go to next line; current line still needs to be processed
break;
case "REFERENCE":
line = ParseReferences(line, ref sequence, stream); // can encounter more than one
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
// don't go to next line; current line still needs to be processed
break;
case "COMMENT":
line = ParseComments(line, ref sequence, stream); // can encounter more than one
metadata = (GenBankMetadata)sequence.Metadata[Helper.GenBankMetadataKey];
// don't go to next line; current line still needs to be processed
break;
case "PRIMARY":
// This header is followed by sequence info in a table format that could be
// stored in a custom object. The first line contains column headers.
// For now, just validate the presence of the headers, and save the data
// as a string.
lineData = GetLineData(line, DataIndent);
tokens = lineData.Split("\t ".ToCharArray(), StringSplitOptions.RemoveEmptyEntries);
// Validating for minimum two headers.
if (tokens.Length != 4)
{
string message = String.Format(
CultureInfo.CurrentCulture,
Properties.Resource.ParserPrimaryLineError,
line);
Trace.Report(message);
throw new InvalidDataException(message);
}
string primaryData = ParseMultiLineData(ref line, Environment.NewLine, DataIndent, stream);
metadata.Primary = primaryData;
// don't go to next line; current line still needs to be processed
break;
// all the following are extracted the same way - possibly multiline
case "DEFINITION":
metadata.Definition = ParseMultiLineData(ref line, " ", DataIndent, stream);
break;
case "ACCESSION":
data = ParseMultiLineData(ref line, " ", DataIndent, stream);
metadata.Accession = new GenBankAccession();
string[] accessions = data.Split(' ');
metadata.Accession.Primary = accessions[0];
for (int i = 1; i < accessions.Length; i++)
{
metadata.Accession.Secondary.Add(accessions[i]);
}
break;
case "DBLINK":
data = ParseMultiLineData(ref line, "\n", DataIndent, stream);
metadata.DbLinks = new List <CrossReferenceLink>();
foreach (string link in data.Split('\n'))
{
tokens = link.Split(':');
if (tokens.Length == 2)
{
CrossReferenceLink newLink = new CrossReferenceLink();
if (string.Compare(tokens[0], CrossReferenceType.Project.ToString(), StringComparison.OrdinalIgnoreCase) == 0)
{
newLink.Type = CrossReferenceType.Project;
}
else if (string.Compare(tokens[0], CrossReferenceType.BioProject.ToString(), StringComparison.OrdinalIgnoreCase) == 0)
{
newLink.Type = CrossReferenceType.BioProject;
}
else
{
newLink.Type = CrossReferenceType.None;
if (string.Compare(tokens[0], TraceAssemblyArchive, StringComparison.OrdinalIgnoreCase) == 0)
{
newLink.Type = CrossReferenceType.TraceAssemblyArchive;
}
}
tokens = tokens[1].Split(',');
for (int i = 0; i < tokens.Length; i++)
{
newLink.Numbers.Add(tokens[i]);
}
metadata.DbLinks.Add(newLink);
}
else
{
ApplicationLog.WriteLine("WARN: unexpected DBLINK header: " + line);
}
}
break;
case "DBSOURCE":
metadata.DbSource = ParseMultiLineData(ref line, " ", DataIndent, stream);
break;
case "KEYWORDS":
metadata.Keywords = ParseMultiLineData(ref line, " ", DataIndent, stream);
break;
case "SEGMENT":
data = ParseMultiLineData(ref line, " ", DataIndent, stream);
const string delimeter = "of";
tokens = data.Split(delimeter.ToCharArray(), StringSplitOptions.RemoveEmptyEntries);
if (tokens.Length == 2)
{
metadata.Segment = new SequenceSegment();
int outvalue;
if (int.TryParse(tokens[0].Trim(), out outvalue))
{
metadata.Segment.Current = outvalue;
}
else
{
ApplicationLog.WriteLine("WARN: unexpected SEGMENT header: " + line);
}
if (int.TryParse(tokens[1].Trim(), out outvalue))
{
metadata.Segment.Count = outvalue;
}
else
{
ApplicationLog.WriteLine("WARN: unexpected SEGMENT header: " + line);
}
}
else
{
ApplicationLog.WriteLine("WARN: unexpected SEGMENT header: " + line);
}
break;
// all the following indicate sections beyond the headers parsed by this method
case "FEATURES":
case "BASE COUNT":
case "ORIGIN":
case "CONTIG":
haveFinishedHeaders = true;
break;
default:
string lineHeader = GetLineHeader(line, DataIndent);
lineData = GetLineData(line, DataIndent);
ApplicationLog.WriteLine(ToString() + "WARN: unknown {0} -> {1}", lineHeader, lineData);
string errMessage = String.Format(
CultureInfo.CurrentCulture,
Properties.Resource.ParseHeaderError,
lineHeader);
Trace.Report(errMessage);
throw new InvalidDataException(errMessage);
}
}
// check for required features
if (!haveParsedLocus)
{
string message = string.Format(CultureInfo.CurrentCulture, Properties.Resource.INVALID_INPUT_FILE, Name);
Trace.Report(message);
throw new InvalidDataException(message);
}
return(line);
}
/// <summary>
/// Private Constructor for clone method.
/// </summary>
/// <param name="other">CrossReferenceLink instance to clone.</param>
private CrossReferenceLink(CrossReferenceLink other)
{
Type = other.Type;
Numbers = new List<string>(other.Numbers);
}