public static Pdb FromPdbid
(string pdbid
, string cachebase // null or @"K:\PdbUnzippeds\$PDBID$.pdb"
, bool?download // null or false
)
{
if (cachebase == null)
{
cachebase = @"K:\PdbUnzippeds\$PDBID$.pdb";
}
if (download == null)
{
download = false;
}
string pdbpath = cachebase.Replace("$PDBID$", pdbid);
if (HFile.Exists(pdbpath))
{
var pdb = Pdb.FromFile(pdbpath);
var last = pdb.elements.Last();
if (last is Pdb.End)
{
return(pdb);
}
if (last.line.Length == 80)
{
return(pdb);
}
if (download.Value)
{ // redownload
pdb = Pdb.FromPdbid(pdbid);
pdb.ToFile(pdbpath);
}
return(pdb);
}
else if (download.Value)
{
Pdb pdb = Pdb.FromPdbid(pdbid);
if (pdb != null)
{
pdb.ToFile(pdbpath);
return(pdb);
}
}
return(null);
}
public static bool GetHessGnmSelfTest()
{
if (HDebug.Selftest() == false)
{
return(true);
}
Pdb pdb = Pdb.FromPdbid("1MJC");
for (int i = 0; i < pdb.atoms.Length; i++)
{
HDebug.Assert(pdb.atoms[0].altLoc == pdb.atoms[i].altLoc);
HDebug.Assert(pdb.atoms[0].chainID == pdb.atoms[i].chainID);
}
List <Vector> coords = pdb.atoms.ListCoord();
double cutoff = 13;
Matlab.Execute("clear");
Matlab.PutMatrix("x", MatrixByArr.FromRowVectorList(coords).ToArray());
Matlab.PutValue("cutoffR", cutoff);
Matlab.Execute(@"% function cx = contactsNew(x, cutoffR)
% Contact matrix within cutoff distance.
% Author: Guang Song
% New: 10/25/2006
%
%n = size(x,1);
% Method 1: slow
%for i=1:n
% center = x(i,:);
% distSqr(:,i) = sum((x-center(ones(n,1),:)).^2,2);
%end
%cx = sparse(distSqr<=cutoffR^2);
% Method 2: fast! about 28 times faster when array size is 659x3
%tot = zeros(n,n);
%for i=1:3
% xi = x(:,ones(n,1)*i);
% %tmp = (xi - xi.').^2;
% %tot = tot + tmp;
% tot = tot + (xi - xi.').^2;
%end
%cx = sparse(tot<=cutoffR^2);
% Method 3: this implementation is the shortest! but sligtly slower than
% method 2
%xn = x(:,:,ones(n,1)); % create n copy x
%xnp = permute(xn,[3,2,1]);
%tot = sum((xn-xnp).^2,2); % sum along x, y, z
%cx = sparse(permute(tot,[1,3,2])<=cutoffR^2);
% put it into one line like below actually slows it down. Don't do that.
%cx = sparse(permute(sum((xn-permute(xn,[3,2,1])).^2,2),[1,3,2])<=cutoffR^2);
%Method 4: using function pdist, which I just know
% this one line implementation is even faster. 2 times than method 2.
cx = sparse(squareform(pdist(x)<=cutoffR));
");
Matlab.Execute(@"% function gnm = kirchhoff(cx)
% the returned gnm provide the kirchhoff matrix
% cx is the contact map.
% Guang Song
% 11/09/06
gnm = diag(sum(cx)) - cx;
");
Matlab.Execute("gnm = full(gnm);");
Matrix gnm_gsong = Matlab.GetMatrix("gnm");
Matlab.Execute("clear;");
Matrix gnm = GetHessGnm(coords.ToArray(), cutoff);
if (gnm_gsong.RowSize != gnm.RowSize)
{
HDebug.Assert(false); return(false);
}
if (gnm_gsong.ColSize != gnm.ColSize)
{
HDebug.Assert(false); return(false);
}
for (int c = 0; c < gnm.ColSize; c++)
{
for (int r = 0; r < gnm.RowSize; r++)
{
if (Math.Abs(gnm_gsong[c, r] - gnm[c, r]) >= 0.00000001)
{
HDebug.Assert(false); return(false);
}
}
}
return(true);
}
