tica
Perform time-independent correlation analysis (TICA)
tica { crdset <COORDS set name>
[mask <mask>] | data <input set arg1> ... }
[lag <time lag>
crdset <COORDS set name> Input coordinates (COORDS) data set.
mask <mask> Selected atoms in input coordinates (COORDS) data set.
data <input set arg1> Input 1D data set name(s), may specify more than once. If any data set is periodic, all need to be periodic.
lag <time lag> TICA lag time in frames.
map {kinetic|commute|none} How to transform the resulting eigenvectors.
kinetic (default) Scale eigenvectors by eigenvalues so that distances in the transformed data approximate kinetic distances; particularly useful if using the projections to cluster.
commute Scale eigenvectors by regularized time scales, sqrt(timescale_i / 2), so that distances in the transformed data will approximate commute distances. Timescales smaller than the lag time are dampened.[Background and reference]
none Do not scale eigenvectors.
name <output set name> Output data set name.
out <file> File to write TICA modes to.
cumvarout <file> File to write eigenvalue cumulative variance to.
Perform time-independent correlation analysis (TICA). Similar to principal component analysis (PCA), TICA calculates eigenvectors/eigenvalues (i.e. eigenmodes) from input data sets (either coordinates or a combination of other 1D data sets). Whereas the eigenvectors from PCA describe the variance in the input data, the eigenvectors from TICA describe the maximal autocorrelation in the input data at the given lag time.[Reference] The analysis can be performed on either coordinates or 1D data sets; the data sets can either be all periodic (in which case they will be converted to cos/sin form) or not.