/** Calculate unambiguous average dihedral angle (in degrees) by converting to
* cartesian coords using x = cos(theta), y = sin(theta), and:
* tan(avgtheta) = avgy / avgx = SUM[sin(theta)] / SUM[cos(theta)]
* See Eq. 2 from Altis et al., J. Chem. Phys., 126 p. 244111 (2007).
*/
static double AvgCalc_Dih( DataSet_1D const& dsIn, ClusterDist::Cframes const& cframesIn,
double& sumx, double& sumy ) {
sumy = 0.0;
sumx = 0.0;
// TODO: Convert angles to radians prior to this call?
for (ClusterDist::Cframes_it frm = cframesIn.begin(); frm != cframesIn.end(); ++frm) {
double theta = dsIn.Dval( *frm ) * Constants::DEGRAD;
sumy += sin( theta );
sumx += cos( theta );
}
return atan2(sumy, sumx) * Constants::RADDEG;
}
int Cluster_ReadInfo::Cluster() {
BufferedLine infile;
if (infile.OpenFileRead( filename_ )) return Err(0);
const char* ptr = infile.Line();
if (ptr == 0) return Err(1);
ArgList infoLine( ptr, " " );
int nclusters = infoLine.getKeyInt("#Clustering:", -1);
if (nclusters == -1) return Err(2);
int nframes = infoLine.getKeyInt("clusters", -1);
if (nframes == -1) return Err(3);
if (nframes != (int)FrameDistances_.Nframes()) {
mprinterr("Error: # frames in cluster info file (%i) does not match"
" current # frames (%zu)\n", nframes, FrameDistances_.Nframes());
return 1;
}
// Scan down to clusters
while (ptr[0] == '#') {
ptr = infile.Line();
if (ptr == 0) return Err(1);
// Save previous clustering info. Includes newline.
if (ptr[1] == 'A' && ptr[2] == 'l' && ptr[3] == 'g')
algorithm_.assign( ptr + 12 ); // Right past '#Algorithm: '
}
// Read clusters
ClusterDist::Cframes frames;
for (int cnum = 0; cnum != nclusters; cnum++) {
if (ptr == 0) return Err(1);
frames.clear();
// TODO: Check for busted lines?
for (int fidx = 0; fidx != nframes; fidx++) {
if (ptr[fidx] == 'X')
frames.push_back( fidx );
}
AddCluster( frames );
mprintf("\tRead cluster %i, %zu frames.\n", cnum, frames.size());
ptr = infile.Line();
}
infile.CloseFile();
mprintf("\tCalculating the distances between each cluster based on centroids.\n");
CalcClusterDistances();
return 0;
}
static double AvgCalc_Std( DataSet_1D const& dsIn, ClusterDist::Cframes const& cframesIn ) {
double val = 0.0;
for (ClusterDist::Cframes_it frm = cframesIn.begin(); frm != cframesIn.end(); ++frm)
val += dsIn.Dval( *frm );
return (val / (double)cframesIn.size());
}
int Cluster_DPeaks::Cluster() {
int err = 0;
// Calculate local densities
if ( useGaussianKernel_ )
err = Cluster_GaussianKernel();
else
err = Cluster_DiscreteDensity();
if (err != 0) return 1;
// Choose points for which the min distance to point with higher density is
// anomalously high.
int nclusters = 0;
if (choosePoints_ == PLOT_ONLY) {
mprintf("Info: Cutoffs for choosing points can be determined visually from the\n"
"Info: density versus min distance to cluster with next highest density file,\n"
"Info: '%s'. Re-run the algorithm with appropriate distancecut and densitycut.\n");
return 0;
} else if (choosePoints_ == MANUAL)
nclusters = ChoosePointsManually();
else
nclusters = ChoosePointsAutomatically();
mprintf("\tIdentified %i cluster centers from density vs distance peaks.\n", nclusters);
// Each remaining point is assigned to the same cluster as its nearest
// neighbor of higher density. Do this recursively until a cluster
// center is found.
int cnum = -1;
for (unsigned int idx = 0; idx != Points_.size(); idx++) {
if (Points_[idx].Cnum() == -1) {// Point is unassigned.
AssignClusterNum(idx, cnum);
//mprintf("Finished recursion for index %i\n\n", idx);
}
}
// Sort by cluster number. NOTE: This invalidates NearestIdx
std::sort( Points_.begin(), Points_.end(), Cpoint::cnum_sort() );
// Determine where each cluster starts and stops in Points array
typedef std::vector<unsigned int> Parray;
Parray C_start_stop;
C_start_stop.reserve( nclusters * 2 );
cnum = -1;
for (Carray::const_iterator point = Points_.begin(); point != Points_.end(); ++point)
{
if (point->Cnum() != cnum) {
if (!C_start_stop.empty()) C_start_stop.push_back(point - Points_.begin()); // end of cluster
C_start_stop.push_back(point - Points_.begin()); // beginning of cluster
cnum = point->Cnum();
}
}
C_start_stop.push_back( Points_.size() ); // end of last cluster
// Noise calculation.
if (calc_noise_) {
mprintf("\tDetermining noise frames from cluster borders.\n");
// For each cluster find a border region, defined as the set of points
// assigned to that cluster which are within epsilon of any other
// cluster.
// NOTE: Could use a set here to prevent duplicate frames.
typedef std::vector<Parray> Barray;
Barray borderIndices( nclusters ); // Hold indices of border points for each cluster.
for (Parray::const_iterator idx0 = C_start_stop.begin();
idx0 != C_start_stop.end(); idx0 += 2)
{
int c0 = Points_[*idx0].Cnum();
//mprintf("Cluster %i\n", c0);
// Check each frame in this cluster.
for (unsigned int i0 = *idx0; i0 != *(idx0+1); ++i0)
{
Cpoint const& point = Points_[i0];
// Look at each other cluster
for (Parray::const_iterator idx1 = idx0 + 2;
idx1 != C_start_stop.end(); idx1 += 2)
{
int c1 = Points_[*idx1].Cnum();
// Check each frame in other cluster
for (unsigned int i1 = *idx1; i1 != *(idx1+1); i1++)
{
Cpoint const& other_point = Points_[i1];
if (FrameDistances_.GetFdist(point.Fnum(), other_point.Fnum()) < epsilon_) {
//mprintf("\tBorder frame: %i (to cluster %i frame %i)\n",
// point.Fnum() + 1, c1, other_point.Fnum() + 1);
borderIndices[c0].push_back( i0 );
borderIndices[c1].push_back( i1 );
}
}
}
}
}
if (debug_ > 0)
mprintf("Warning: Cluster numbers here may not match final cluster numbers.\n"
"\tBorder Frames:\n");
for (Parray::const_iterator idx = C_start_stop.begin();
idx != C_start_stop.end(); idx += 2)
{
int c0 = Points_[*idx].Cnum();
if (debug_ > 0)
mprintf("\tCluster %u: %u frames: %u border frames:", c0, *(idx+1) - *idx,
borderIndices[c0].size());
if (borderIndices[c0].empty()) {
if (debug_ > 0) mprintf(" No border points.\n");
} else {
int highestDensity = -1;
// Find highest density in border region.
for (Parray::const_iterator bidx = borderIndices[c0].begin();
bidx != borderIndices[c0].end(); ++bidx)
{
if (highestDensity == -1)
highestDensity = Points_[*bidx].PointsWithinEps();
else
highestDensity = std::max(highestDensity, Points_[*bidx].PointsWithinEps());
if (debug_ > 0) mprintf(" %i", Points_[*bidx].Fnum()+1);
}
if (debug_ > 0) mprintf(". Highest density in border= %i\n", highestDensity);
// Mark any point with density <= highest border density as noise.
for (unsigned int i = *idx; i != *(idx+1); i++)
{
Cpoint& point = Points_[i];
if (point.PointsWithinEps() <= highestDensity) {
point.SetCluster( -1 );
if (debug_ > 1)
mprintf("\t\tMarking frame %i as noise (density %i)\n",
point.Fnum()+1, point.PointsWithinEps());
}
}
}
}
}
// Add the clusters.
for (Parray::const_iterator idx = C_start_stop.begin();
idx != C_start_stop.end(); idx += 2)
{
ClusterDist::Cframes frames;
for (unsigned int i = *idx; i != *(idx+1); i++) {
if (Points_[i].Cnum() != -1)
frames.push_back( Points_[i].Fnum() );
}
if (!frames.empty())
AddCluster( frames );
}
// Calculate the distances between each cluster based on centroids
CalcClusterDistances();
return 0;
}
