Divide-and-Conquer Multiple Sequence Alignment Parameters of DCA

In general, DCA takes as input a family of sequences and a matrix of pairwise letter distances as well as gap initiation and gap extension parameters.

• Sequences
  DCA accepts protein, DNA, RNA, or any ASCII sequences. Sequences can be either typed in manually in the provided lines (e.g. by dragging and dropping with the computer mouse), or they can be loaded from a local sequence file. Accepted input formats are FASTA and GDE. For an example see here.

• Substitution matrix
  You can either use any of the provided substitution matrices (PAM250, PAM160, Blosum30, Blosum45, Blosum62, Gonnet250, Gonnet120, DNA/RNA, or unit cost) which are displayed upon clicking the "show matrix" button in the submission form, or you can provide your own (distance) substitution matrix in MSA format.

• Gap parameters
  The affine cost of a gap is computed by the formula $g(l) = \alpha + \beta l$ where $\alpha$ is the "gap initiation cost" and $\beta$ is the "gap extension cost".

• Free shift
  DCA by default does not penalize gaps at either end of the sequences to make the compensation of differences in the length of the seuences free of charge. This free shift option can be deactivated.

• Approximate cut positions/FDCA
  The last, most time consuming phase of the search for cut positions can be deactivated so that the sequences are cut at approximate slicing positions, which generally yields slightly less accurate but often much faster alignments. A more decent explanation of this heuristic is found here.

• Recursion stop size
  The recursion stop size can be set to any number $L \geq 1$. Of course, too large an L (e.g. $L \geq 100$) can result in very long running times and very big memory usage due to the resulting MSA-runs. On the other hand, too small an L (e.g. $L < 5$) can result in empty subsequences at the end of the iteration which may lead to bad alignments.

• Window size
  To correct the alignment in the proximity of division sites, the sequences can be re-aligned inside a window of size $W \geq 0$placed across each slicing site.

• Weight intensity
  The weight intensity $\lambda$can be set to any value between 0 (no weighting) and 1.0 (max. weighting). Here is the formula.