internal static HlaIGenotype Union(params HlaIGenotype[] hlaIGenotypes)
{
HlaI A1 = new HlaI(HlaILocus.A);
HlaI A2 = new HlaI(HlaILocus.A);
HlaI B1 = new HlaI(HlaILocus.B);
HlaI B2 = new HlaI(HlaILocus.B);
HlaI C1 = new HlaI(HlaILocus.C);
HlaI C2 = new HlaI(HlaILocus.C);
foreach (var geno in hlaIGenotypes)
{
A1.ParseInto(geno.AAlleles.Item1.ToString());
A2.ParseInto(geno.AAlleles.Item2.ToString());
B1.ParseInto(geno.BAlleles.Item1.ToString());
B2.ParseInto(geno.BAlleles.Item2.ToString());
C1.ParseInto(geno.CAlleles.Item1.ToString());
C2.ParseInto(geno.CAlleles.Item2.ToString());
}
var result = new HlaIGenotype(A1, A2, B1, B2, C1, C2);
return(result);
}
protected override void WriteDistilled(IEnumerable <KeyValuePair <string, List <HlaIGenotype> > > patientsAndGenotypes, bool multipleGenotypesPerPatient)
{
var patientsAndGenotypesAsList = patientsAndGenotypes.ToList();
Dictionary <string, MultinomialStatistics> pidToMultinomial = new Dictionary <string, MultinomialStatistics>();
Dictionary <string, HlaIGenotype> pidGenoNameToGeno = new Dictionary <string, HlaIGenotype>(patientsAndGenotypesAsList.Count);
foreach (var pidAndGenos in patientsAndGenotypes)
{
if (pidAndGenos.Value.Count == 1)
{
pidGenoNameToGeno.Add(pidAndGenos.Key, pidAndGenos.Value.Single());
}
else
{
pidToMultinomial.Add(pidAndGenos.Key, MultinomialStatistics.GetInstance(pidAndGenos.Value.Select(g => g.Probability))); // keep track of the probabilities in this dictionary
HlaIGenotype unionGenotype = HlaIGenotype.Union(pidAndGenos.Value.ToArray());
pidGenoNameToGeno.Add(pidAndGenos.Key, unionGenotype); //Add the union genotype to the matrix.
for (int i = 0; i < pidAndGenos.Value.Count; i++)
{
string pidGenoId = pidAndGenos.Key + "_" + i;
pidGenoNameToGeno.Add(pidGenoId, pidAndGenos.Value[i]);
}
}
}
Matrix <string, string, SufficientStatistics> m = DenseMatrix <string, string, SufficientStatistics> .CreateDefaultInstance(
HlaEnumerator.GenerateHlas(patientsAndGenotypes.SelectMany(pidAndGenos => pidAndGenos.Value)),
pidGenoNameToGeno.Keys,
MissingStatistics.GetInstance());
foreach (string hlastr in m.RowKeys)
{
HlaI hla = HlaI.Parse(hlastr);
foreach (string pid in m.ColKeys)
{
bool?match = pidGenoNameToGeno[pid].Matches(hla, MixtureSemantics);
m.SetValueOrMissing(hlastr, pid, BooleanStatistics.GetInstance(match));
//m.SetValueOrMissing(hlastr, pid, !match.HasValue ? "?" : match.Value ? "1" : "0");
}
}
m.WriteDense(this.Out.CreateTextOrUseConsole());
string baseName = Path.GetFileNameWithoutExtension(this.Out.ToString());
string probabilityFile = this.Out.ToString() == "-" ? "HaplotypeCompletionProbs.txt" : this.Out.ToString().Replace(baseName, baseName + "_haplotypeProbs");
pidToMultinomial.WriteDelimitedFile(probabilityFile);
}
/// <summary>
/// Returns true if and only if we know for sure that this genotype matches the given Hla. Returns false if an only if we
/// know for sure that this genotype does not describe the given hla. Otherwise, returns null.
/// </summary>
/// <param name="unambiguousHla"></param>
/// <returns></returns>
public bool?Matches(HlaI unambiguousHla, MixtureSemantics mixtureSemantics = MixtureSemantics.Uncertainty)
{
Helper.CheckCondition <ArgumentException>(!unambiguousHla.IsAmbiguous, "Can only check if you have an uynambiguous Hla");
Tuple <HlaI, HlaI> locus;
switch (unambiguousHla.Locus)
{
case HlaILocus.A:
locus = AAlleles; break;
case HlaILocus.B:
locus = BAlleles; break;
case HlaILocus.C:
locus = CAlleles; break;
default:
throw new Exception("Can't get here.");
}
UOPair <HlaI> locusToCompare = UOPair.Create(locus.Item1, locus.Item2); // easier to treat as unordered pair here, since it will order the nulls.
if (locusToCompare.First == null && locusToCompare.Second == null)
{
return(null); // we know nothing about this locus
}
// note: if one of them is null, it's first. so start with second.
bool?secondIsOther = locusToCompare.Second.Matches(unambiguousHla, mixtureSemantics);
if (secondIsOther.HasValue && secondIsOther.Value)
{
return(true); // if second is a match, then we know it has it.
}
if (locusToCompare.First == null)
{
return(null); //don't know anything about first.
}
bool?firstIsOther = locusToCompare.First.Matches(unambiguousHla, mixtureSemantics);
if (firstIsOther.HasValue && firstIsOther.Value)
{
return(true); // first is a match, so we know we have it
}
if (!firstIsOther.HasValue || !secondIsOther.HasValue)
{
return(null); // neither has it, so if either is missing, we don't know
}
return(false); // if we get here, then both alleles reported false.
}
private static HlaI ParseHla(HlaILocus locus, string hlaString, HlaI otherCopyAtSameLocusOrNull = null)
{
Helper.CheckCondition <ParseException>(!string.IsNullOrWhiteSpace(hlaString), "Blank entries are not allowed. Use ? for missing or - for homozygous (second column of locus only).");
switch (hlaString)
{
case "?":
return(null);
case "-":
Helper.CheckCondition <ParseException>(otherCopyAtSameLocusOrNull != null, "Can't mark the first column as homozygous. Only A2, B2 or C2 can be marked with -.");
return(otherCopyAtSameLocusOrNull);
default:
HlaI hla = new HlaI(locus);
hla.ParseInto(hlaString);
return(hla);
}
}
public HlaIGenotype(HlaI A1, HlaI A2, HlaI B1, HlaI B2, HlaI C1, HlaI C2)
{
AAlleles = Tuple.Create(A1, A2);
BAlleles = Tuple.Create(B1, B2);
CAlleles = Tuple.Create(C1, C2);
}
public virtual Dictionary <string, List <HlaIGenotype> > ParseFile()
{
Dictionary <string, List <HlaIGenotype> > result = new Dictionary <string, List <HlaIGenotype> >();
using (TextReader reader = HlaFile.OpenTextOrUseConsole(stripComments: true))
{
string header = reader.ReadLine();
bool isHlaCompletionFile = header.Equals(HlaCompletionHeader);
int colCount = header.Split('\t').Length;
Helper.CheckCondition <ParseException>(isHlaCompletionFile || colCount == 7 || colCount == 8, "Expect a file with 7 or 8 columns. Read {0}", header);
bool parseProbabilityLine = isHlaCompletionFile || colCount == 8;
string line;
while (null != (line = reader.ReadLine()))
{
string[] fields = line.Split('\t');
try
{
string pid = fields[0];
HlaI a1 = ParseHla(HlaILocus.A, fields[1]);
HlaI a2 = ParseHla(HlaILocus.A, fields[2], a1);
HlaI b1 = ParseHla(HlaILocus.B, fields[3]);
HlaI b2 = ParseHla(HlaILocus.B, fields[4], b1);
HlaI c1 = ParseHla(HlaILocus.C, fields[5]);
HlaI c2 = ParseHla(HlaILocus.C, fields[6], c1);
HlaIGenotype genotype = new HlaIGenotype(a1, a2, b1, b2, c1, c2);
genotype.SetMaxResolution(MaxResolution);
if (parseProbabilityLine)
{
double p;
(double.TryParse(fields[7], out p) && p >= 0 && p <= 1).Enforce <ParseException>("The 8th field is not a valid probability. Read {0}.", p);
genotype.Probability = p;
}
result.GetValueOrDefault(pid).Add(genotype);
}
catch (ParseException p)
{
throw new ParseException(p.Message + " for patient" + fields[0]);
}
}
}
if (MaxResolution == HlaIResolution.Group)
{
foreach (var pidAndGenos in result)
{
if (pidAndGenos.Value.Count > 1)
{
// we've stripped out some of the 4 digits. The resulting list may contain duplicates. Collapse those together and sum up their probabilities.
var collapsedGenotypes = from geno in pidAndGenos.Value
let hlaString = geno.EnumerateAll().StringJoin(" ")
group geno by hlaString into identicalHaps
select new HlaIGenotype(
identicalHaps.First().AAlleles.Item1,
identicalHaps.First().AAlleles.Item2,
identicalHaps.First().BAlleles.Item1,
identicalHaps.First().BAlleles.Item2,
identicalHaps.First().CAlleles.Item1,
identicalHaps.First().CAlleles.Item2)
{
Probability = identicalHaps.Sum(h => h.Probability)
};
pidAndGenos.Value.Clear();
pidAndGenos.Value.AddRange(collapsedGenotypes);
}
}
}
return(result);
}
