Bonsai uses a position-specific score matrix (PSSM) for multiple alignments. The PSSM frame displays these values for an aligned profile, and permits editing of some of these values. WARNING: edit the PSSM cautiously!
Column Weight: the weight that a particular aligned column is given when used in alignments. When a profile is constructed de novo, this value is the same for all columns (1.0). If you want to change the importance of this column for future alignments, edit this value. For example, you may know that specific regions of an alignment are biologically critical or dispensable, and you may change the weight of those columns up or down to account for this. Since a de novo profile construction can't know about these values ahead of time, you must edit these manually.
Scores: each amino acid will receive a specific integer score when aligned against this column. When a profile is constructed, the values of these scores are determined by the score matrix used and the weighted counts of amino acids in this column.
Weighted Counts: each amino acid in a column of a profile is present a fixed number of times. This count is multiplied by the weights of the sequences containing those amino acids to give rise to a weighted residue count. This value is currently not editable in Bonsai (and it probably shouldn't ever be). It is displayed for information only. Sequence weights were computed during profile construction according to the guide tree by the algorithm of [ref]. These weights are important to avoid overrepresenting closely related sequences. Since available sequence databases are heavily biased towards related organisms (bacteria and mammals primarily) this correction is often critical unless the user carefully edits the sequence set used to construct the profile.
Expanded Warning: the values in an unedited PSSM have considerable justification in the probabilistic model of the profile. You may possess information that permits you to usefully change these values, but in doing so you abandon the probabilistic model to some degree. It is generally best to make such changes with caution and change the values only moderately. Even modest changes (e.g. doubling the weight of a column) may have large effects on subsequent alignments. I don't know of any rules of thumb here, so you may want to try small changes and experiment with the alignments your new PSSM produces.
You want to align new sequences to the extracellular domain of a protein sequence family. You know from experimental results that the critical determinant for this domain is a set of highly conserved cysteines. Pairs of these cysteines form invariant disulphide bonds that define a stable set of folds. The remaining amino acids are more variable in spacing and identity because they define the ligand-binding specificity for specific members of the family. You have selected a representative set of these proteins and constructed a multiple alignment profile for them, perhaps even editing the gaps slightly to force the critical cysteines to align (if they don't already). You edit the PSSM for this profile to enhance the weight of these Cys columns, encouraging future alignments to match these Cys residues. If your profile has nothing but Cys residues at these positions, you should probably leave the score values alone, but if there are rare exceptions you may also want to edit the scores up or down.