How can I make feature vectors using PSSM matrix?

Question

[MT]

How can I make feature vectors from PSSM matrix ? For example I know a simple way which is calculating sum of all the numbers in a row of the matrix, this way we will have a single number for each row ( in fact for each amino acid) which is not dependent on L ( length ) and finally we will have a vector of 20 numbers ( for the 20 amino acids) and we can use it for machine learning purposes. Now I need some more similar ways to produce such vectors from PSSM matrix.

[Mike]

I think in case of PSSM, the vector matrix should be 400 (20*20), i.e 20 values for each amino acid.

[MT]

PSSM is actually 20*L (20 rows for the 20 amino acids, and L column for different positions), and we want to extract a feature vector (20 rows * 1 column) from it.