Matrix is singular

Daniel Álvarez García · Post by **Daniel Álvarez García** » Fri Mar 22, 2013 8:22 am

Dear users and developers,

I'm trying to run EMMA for a ligand-protein binding event (for a very small organic molecule) for which I have several trajectories of 20ns each. After clustering, assigning snapshots to clusters and estimating the lagtime, if I estimate the transition matrix using -reversible as argument, all runs fine but if I do the same with out -reversible estimation the program mm_estimate yields the following error:

Code: Select all

Exception in thread "main" java.lang.IllegalArgumentException: Matrix is singular.
	at cern.colt.matrix.linalg.LUDecompositionQuick.solve(Unknown Source)
	at cern.colt.matrix.linalg.LUDecomposition.solve(Unknown Source)
	at cern.colt.matrix.linalg.Algebra.solve(Unknown Source)
	at cern.colt.matrix.linalg.Algebra.inverse(Unknown Source)
	at sherlock.emma.Pcca_Simplex.perform(Pcca_Simplex.java:205)
	at sherlock.emma.cmd.PccaCmd.run(PccaCmd.java:177)
	at de.martin_senne.jcommandline.AbstractCmd.check(AbstractCmd.java:33)
	at de.martin_senne.jcommandline.AbstractCmd.go(AbstractCmd.java:45)
	at sherlock.emma.cmd.PccaCmd.main(PccaCmd.java:282)

How can this happen to obtain a singular transition matrix? Is it due to insufficient sampling?
When should one use the -reversible argument?

Thanks for the help!

Daniel