public SearchableDatabase(FastaDatabase fastaDatabase)
{
FastaDatabase = fastaDatabase;
_sequence = fastaDatabase.GetSequence();
_suffixArray = new int[_sequence.Length];
SAIS.sufsort(_sequence, _suffixArray, _sequence.Length);
var neighboringLcps = new byte[_suffixArray.Length];
neighboringLcps[0] = 0;
for (var i = 1; i < _suffixArray.Length; i++)
{
var lcp = IndexedDatabase.GetLcp(_sequence, _suffixArray[i - 1], _suffixArray[i]);
neighboringLcps[i] = lcp;
}
_leftLcps = new byte[_suffixArray.Length];
_rightLcps = new byte[_suffixArray.Length];
InitializeLcps(neighboringLcps, _leftLcps, _rightLcps, 0, _suffixArray.Length-1);
}
/// <summary>
/// Constructor
/// </summary>
/// <param name="fastaDatabase"></param>
public SearchableDatabase(FastaDatabase fastaDatabase)
{
FastaDatabase = fastaDatabase;
_sequence = fastaDatabase.GetSequence();
_suffixArray = new int[_sequence.Length];
SAIS.sufsort(_sequence, _suffixArray, _sequence.Length);
var neighboringLcps = new byte[_suffixArray.Length];
neighboringLcps[0] = 0;
for (var i = 1; i < _suffixArray.Length; i++)
{
var lcp = IndexedDatabase.GetLcp(_sequence, _suffixArray[i - 1], _suffixArray[i]);
neighboringLcps[i] = lcp;
}
_leftLcps = new byte[_suffixArray.Length];
_rightLcps = new byte[_suffixArray.Length];
InitializeLcps(neighboringLcps, _leftLcps, _rightLcps, 0, _suffixArray.Length - 1);
}
private void CreatePermutedLongestCommonPrefixFile()
{
if (File.Exists(_pLcpFilePath))
{
File.Delete(_pLcpFilePath);
}
var sequence = FastaDatabase.GetSequence();
//Console.WriteLine("Annotation: {0}", System.Text.Encoding.ASCII.GetString(sequence));
var suffixArray = new int[sequence.Length - 1];
SAIS.sufsort(sequence, suffixArray, sequence.Length - 1);
var prevIndex = sequence.Length - 1;
var pLcp = new byte[suffixArray.Length];
// Data dependency: cannot run in parallel
foreach (var index in suffixArray)
{
var lcp = GetLcp(sequence, prevIndex, index);
pLcp[index] = lcp;
//Console.WriteLine("{0}\t{1}", System.Text.Encoding.ASCII.GetString(sequence, offset, sequence.Length - offset - 1), lcp);
prevIndex = index;
}
using (var fs = new FileStream(_pLcpFilePath, FileMode.OpenOrCreate, FileAccess.Write))
{
foreach (var lcp in pLcp)
{
//Console.WriteLine("LCP: {0}", lcp);
fs.WriteByte(lcp);
}
fs.Write(BitConverter.GetBytes(FastaDatabase.FileFormatId), 0, sizeof(int));
fs.Write(BitConverter.GetBytes(FastaDatabase.GetLastWriteTimeHash()), 0, sizeof(int));
}
}
public void CountMatchedScansPerProtein()
{
var methodName = MethodBase.GetCurrentMethod().Name;
TestUtils.ShowStarting(methodName);
const int minTagLength = 6;
var proteinToScan = new Dictionary<string, HashSet<int>>();
const string fastaFilePath = @"D:\Research\Data\CommonContaminants\H_sapiens_Uniprot_SPROT_2013-05-01_withContam.fasta";
if (!File.Exists(fastaFilePath))
{
Assert.Ignore(@"Skipping test {0} since file not found: {1}", methodName, fastaFilePath);
}
var fastaDb = new FastaDatabase(fastaFilePath);
var searchableDb = new SearchableDatabase(fastaDb);
Console.WriteLine(@"Sequence length: {0}", fastaDb.GetSequence().Length);
//const string tagFilePath = @"H:\Research\QCShew_TopDown\Production\QC_Shew_Intact_26Sep14_Bane_C2Column3_seqtag.tsv";
//const string tagFilePath = @"\\protoapps\UserData\Jungkap\Co_culture\23B_pellet_TD_3Feb14_Bane_PL011402.seqtag";
const string tagFilePath = @"D:\MassSpecFiles\co_culture\23A_pellet_TD_3Feb14_Bane_PL011402.seqtag";
if (!File.Exists(tagFilePath))
{
Assert.Ignore(@"Skipping test {0} since file not found: {1}", methodName, tagFilePath);
}
var isHeader = true;
var numMatchedPairs = 0;
foreach (var line in File.ReadAllLines(tagFilePath))
{
if (isHeader)
{
isHeader = false;
continue;
}
var token = line.Split('\t');
if (token.Length != 3) continue;
var scan = Convert.ToInt32(token[0]);
var tag = token[1];
if (tag.Length < minTagLength) continue;
foreach (var matchedProtein in searchableDb.FindAllMatchedSequenceIndices(tag)
.Select(index => fastaDb.GetProteinName(index)))
{
++numMatchedPairs;
HashSet<int> matchedScans;
if (proteinToScan.TryGetValue(matchedProtein, out matchedScans))
{
matchedScans.Add(scan);
}
else
{
matchedScans = new HashSet<int> {scan};
proteinToScan.Add(matchedProtein, matchedScans);
}
}
}
var numMatchedProteins = proteinToScan.Keys.Count;
var numAllProteins = fastaDb.GetNumEntries();
Console.WriteLine("NumAllProteins: {0}", numAllProteins);
Console.WriteLine("NumMatchedProteins: {0}", numMatchedProteins);
Console.WriteLine("AvgMatchedScansPerProtein: {0}", numMatchedPairs / (float)numAllProteins);
}
public void CountMatchedProteins()
{
var methodName = MethodBase.GetCurrentMethod().Name;
TestUtils.ShowStarting(methodName);
const int minTagLength = 3;
var scanToProtein = new Dictionary<int, string>();
var idTag = new Dictionary<int, bool>();
const string resultFilePath = @"H:\Research\ProMex\QC_Shew_Intact_26Sep14_Bane_C2Column3_IcTda.tsv";
if (!File.Exists(resultFilePath))
{
Assert.Ignore(@"Skipping test {0} since file not found: {1}", methodName, resultFilePath);
}
var parser = new TsvFileParser(resultFilePath);
var scans = parser.GetData("Scan").Select(s => Convert.ToInt32(s)).ToArray();
var proteinNames = parser.GetData("ProteinName").ToArray();
var qValues = parser.GetData("QValue").Select(Convert.ToDouble).ToArray();
for (var i = 0; i < qValues.Length; i++)
{
if (qValues[i] > 0.01) break;
scanToProtein.Add(scans[i], proteinNames[i]);
idTag.Add(scans[i], false);
}
const string rawFilePath = @"H:\Research\QCShew_TopDown\Production\QC_Shew_Intact_26Sep14_Bane_C2Column3.raw";
if (!File.Exists(rawFilePath))
{
Assert.Ignore(@"Skipping test {0} since file not found: {1}", methodName, rawFilePath);
}
var run = PbfLcMsRun.GetLcMsRun(rawFilePath);
const string fastaFilePath = @"H:\Research\QCShew_TopDown\Production\ID_002216_235ACCEA.fasta";
if (!File.Exists(fastaFilePath))
{
Assert.Ignore(@"Skipping test {0} since file not found: {1}", methodName, fastaFilePath);
}
// const string fastaFilePath = @"H:\Research\QCShew_TopDown\Production\ID_002216_235ACCEA.icsfldecoy.fasta";
// const string fastaFilePath =
// @"D:\Research\Data\CommonContaminants\H_sapiens_Uniprot_SPROT_2013-05-01_withContam.fasta";
var fastaDb = new FastaDatabase(fastaFilePath);
var searchableDb = new SearchableDatabase(fastaDb);
Console.WriteLine("Sequence length: {0}", fastaDb.GetSequence().Length);
const string tagFilePath = @"H:\Research\QCShew_TopDown\Production\QC_Shew_Intact_26Sep14_Bane_C2Column3.seqtag";
if (!File.Exists(tagFilePath))
{
Assert.Ignore(@"Skipping test {0} since file not found: {1}", methodName, tagFilePath);
}
var hist = new Dictionary<int, int>();
var scanSet = new HashSet<int>();
HashSet<string> proteinSetForThisScan = null;
var prevScan = -1;
var totalNumMatches = 0L;
var isHeader = true;
foreach (var line in File.ReadAllLines(tagFilePath))
{
if (isHeader)
{
isHeader = false;
continue;
}
var token = line.Split('\t');
if (token.Length < 3) continue;
var scan = Convert.ToInt32(token[0]);
var proteinId = scanToProtein.ContainsKey(scan) ? scanToProtein[scan] : null;
if (scan != prevScan)
{
if (proteinSetForThisScan != null)
{
var numMatches = proteinSetForThisScan.Count;
int numOcc;
if (hist.TryGetValue(numMatches, out numOcc)) hist[numMatches] = numOcc + 1;
else hist.Add(numMatches, 1);
}
prevScan = scan;
proteinSetForThisScan = new HashSet<string>();
}
scanSet.Add(scan);
var tag = token[1];
if (tag.Length < minTagLength) continue;
if (proteinSetForThisScan == null) continue;
var numMatchesForThisTag = 0;
foreach (var matchedProtein in searchableDb.FindAllMatchedSequenceIndices(tag)
.Select(index => fastaDb.GetProteinName(index)))
{
proteinSetForThisScan.Add(matchedProtein);
++numMatchesForThisTag;
if (proteinId != null && matchedProtein.Equals(proteinId))
{
idTag[scan] = true;
}
}
totalNumMatches += numMatchesForThisTag;
// if (numMatchesForThisTag > 10)
// {
// Console.WriteLine("{0}\t{1}", tag, numMatchesForThisTag);
// }
}
if (proteinSetForThisScan != null)
{
var numMatches = proteinSetForThisScan.Count;
int numOcc;
if (hist.TryGetValue(numMatches, out numOcc)) hist[numMatches] = numOcc + 1;
else hist.Add(numMatches, 1);
}
Console.WriteLine("AvgNumMatches: {0}", totalNumMatches/(float)scanSet.Count);
Console.WriteLine("Histogram:");
foreach (var entry in hist.OrderBy(e => e.Key))
{
Console.WriteLine("{0}\t{1}", entry.Key, entry.Value);
}
Console.WriteLine("NumId: {0}", idTag.Count);
Console.WriteLine("NumIdByTag: {0}", idTag.Select(e => e.Value).Count(v => v));
}
