private void doTraining(CreatePredictorVM vm, IEnumerable <PDBFile> trainingFiles, IDictionary <string, IDictionary <Residue, bool> > referenceInterface, IDictionary <string, IDictionary <string, ProteinGraph> > graphs)
{
var Feature = new AverageHydrophobicityFeature();
var values = new LinkedList <AgO <double, ResidueNode> >();
foreach (var trainingFile in trainingFiles)
{
var req = new RequestRasa(trainingFile);
req.RequestInDefaultContext();
var rasa = req.Rasavalues;
foreach (var entry in graphs[trainingFile.Name])
{
Feature.Compute(entry.Value);
//values.AddRange(entry.Value.Nodes.Select(node => new AgO<double, ResidueNode>(Feature.ValueOf(node.Data.Residue) * rasa[node.Data.Residue], node.D
//)));
}
}
Max = values.Max(v => v.Data1);
//var intervals = values.DivideByScoreAequidistant(val => val.Data1, vm.DivisionIntervals);
//var pts = new LinkedList<AgO<double, double>>();
//for (int i = 0; i < intervals.Length; i++)
//{
// var interval = intervals[i];
// var pt = new AgO<double, double>();
// pt.Data1 = ((double)i + 1) / intervals.Length;
// var iface = (interval.Sum(val => referenceInterface[val.Data2.Residue.PDB.Name][val.Data2.Residue] ? 1.0 : 0.0));
// pt.Data2 = iface / interval.Count();
// pts.Add(pt);
//}
//ValueFunction = new StepFunction(pts);
}
private void OnRequest(RequestRasa obj)
{
if (buffer.ContainsKey(obj.File))
{
obj.Rasavalues = new Dictionary <Residue, double>(buffer[obj.File]);
}
else
{
foreach (var chain in obj.File.Chains)
{
var rasaFile = Directory.GetFiles(RasaFiles).FirstOrDefault(file => file.Contains(obj.File.Name + "_" + chain.Name));
if (rasaFile == null)
{
var temp = " -jar protein-pdb-asa.jar -pdbfile "
+ PDBFiles + obj.File.Name + "_" + chain.Name + ".txt" + " -pdb " + obj.File.Name + " -chain " + chain.Name + " -rasafile "
+ RasaFiles + obj.File.Name + "_" + chain.Name + ".rasa";
//compute Rasa:
var processInfo = new ProcessStartInfo("java", temp)
{
CreateNoWindow = true,
UseShellExecute = false
};
Process proc;
if ((proc = Process.Start(processInfo)) == null)
{
throw new InvalidOperationException("??");
}
proc.WaitForExit();
int exitCode = proc.ExitCode;
proc.Close();
}
//else
{
rasaFile = rasaFile ?? Directory.GetFiles(RasaFiles).FirstOrDefault(file => file.Contains(obj.File.Name + "_" + chain.Name));
using (var reader = new StreamReader(rasaFile))
{
string line = "";
while ((line = reader.ReadLine()) != null)
{
var words = Regex.Split(line, "\\s");
var nodeid = words[0].Substring(7);
var rasa = Math.Min(1.0, double.Parse(words[3], CultureInfo.InvariantCulture));
var residue = chain.Residues.First(res => res.Id.Equals(nodeid));
obj.Rasavalues.Add(residue, rasa);
}
}
}
}
buffer.Add(obj.File, new Dictionary <Residue, double>(obj.Rasavalues));
}
}
private Dictionary <string, Dictionary <Residue, bool> > doPrediction(PDBFile file, IDictionary <string, ProteinGraph> graphs)
{
var dict = new Dictionary <string, Dictionary <Residue, bool> >();
var Feature = new AverageHydrophobicityFeature();
var req = new RequestRasa(file);
req.RequestInDefaultContext();
var rasa = req.Rasavalues;
//Threshold = 0.5;
//foreach (var graph in graphs)
//{
// Feature.Compute(graph.Value);
// var valdict = new Dictionary<ResidueNode, double[]>();
// var edgeScores = new double[2, 2] { { 1.2, 0.6 }, { 0.6, 1.0 } };
// foreach (var node in graph.Value.Nodes)
// {
// var val = (Feature.ValueOf(node.Data.Residue) * rasa[node.Data.Residue]) / Max;
// valdict.Add(node, new double[2] { val, 1 - val });
// }
// var edgedict = new Dictionary<SimpleEdge<ResidueNode>, double[,]>();
// foreach (var edge in graph.Value.Edges)
// {
// edgedict.Add(edge, edgeScores);
// }
// var predDict = new Dictionary<Residue, bool>();
// dict.Add(graph.Key, predDict);
// var crfGraph = graph.Value.CreateGraph(valdict, edgedict);
// //var options = new MarginalizeByResamplingOptions(2, 200);
// //foreach (var node in crfNodes)
// //{
// // var probs = MarginalizeByResampling.Do(crfNodes, node, options);
// // predDict.Add(graph.Value.Nodes.First(nd => nd.Residue.Id.Equals(node.Id)).Residue, probs[0] >= Threshold);
// //}
// var result = SimpleSampling.Do(crfGraph, 2000);
// foreach (var node in crfGraph.Nodes)
// {
// var score1 = result[new Assign() { Node = node, Label = 0 }];
// var score2 = result[new Assign() { Node = node, Label = 1 }];
// var prob = score2 / (score1 + score2);
// predDict.Add(graph.Value.Nodes.First(nd => nd.Data.Residue.Id.Equals(node.Data.Id)).Residue, prob >= Threshold);
// }
//}
return(dict);
}
private void OnProvidePDBInfo(ProvidePDBInfo obj)
{
var options = obj.Options;
var pdbId = options.PDBId;
var pdbinfo = default(PDBInfo);
var req = new GetPDBs(pdbId.ToIEnumerable());
req.RequestInDefaultContext();
var file = req.Files.FirstOrDefault();
if (req.Files.NotNullOrEmpty())
{
pdbinfo = new PDBInfo(file);
}
if (options.ComputeRasa)
{
var reqR = new RequestRasa(file);
reqR.RequestInDefaultContext();
var rasa = reqR.Rasavalues;
}
if (options.ComputeInterface)
{
var request = new RequestInterface(pdbinfo.PDBFile.ToIEnumerable(), options.InterfaceDefDistance, options.InterfaceUseVanDerWaalsRadii);
request.RequestInDefaultContext();
pdbinfo.Interface.Clear();
foreach (var residue in pdbinfo.PDBFile.Residues)
{
pdbinfo.Interface.Add(residue, request.Interface[pdbinfo.PDBFile.Name][residue]);
}
}
if (options.ComputeGraph)
{
pdbinfo.Graphs = CreateProteinGraph.Do(pdbinfo.PDBFile, options.GraphNeighbourDistance, options.GraphNodeCenterDef);
if (!(pdbinfo.Interface.Count == 0))
{
foreach (var pgraph in pdbinfo.Graphs)
{
pgraph.Value.Nodes.Where(n => pdbinfo.Interface[n.Data.Residue]).Each((node) => node.Data.IsRefInterface = true);
}
}
}
obj.PDBInfo = pdbinfo;
}
private void doTraining(CreatePredictorVM vm, IEnumerable <PDBFile> trainingFiles, IDictionary <string, IDictionary <Residue, bool> > referenceInterface, IDictionary <string, IDictionary <string, ProteinGraph> > graphs)
{
var Feature = new AverageHydrophobicityFeature();
var values = new LinkedList <AgO <double, ResidueNode> >();
foreach (var trainingFile in trainingFiles)
{
var req = new RequestRasa(trainingFile);
req.RequestInDefaultContext();
var rasa = req.Rasavalues;
foreach (var entry in graphs[trainingFile.Name])
{
Feature.Compute(entry.Value);
//values.AddRange(entry.Value.Nodes.Select(node => new AgO<double, ResidueNode>(Feature.ValueOf(node.Data.Residue) * rasa[node.Data.Residue], node)));
}
}
Max = values.Max(v => v.Data1);
}
private Dictionary <string, Dictionary <Residue, bool> > doPrediction(PDBFile file, IDictionary <string, ProteinGraph> graphs)
{
var dict = new Dictionary <string, Dictionary <Residue, bool> >();
var Feature = new AverageHydrophobicityFeature();
var req = new RequestRasa(file);
req.RequestInDefaultContext();
var rasa = req.Rasavalues;
foreach (var graph in graphs)
{
Feature.Compute(graph.Value);
var predDict = new Dictionary <Residue, bool>();
dict.Add(graph.Key, predDict);
foreach (var node in graph.Value.Nodes)
{
predDict.Add(node.Data.Residue, (Feature.ValueOf(node.Data.Residue) * rasa[node.Data.Residue]) / Max >= Threshold);
}
}
return(dict);
}
