GWW makes an excellent point about your code design. For learning purposes, it's best to post what code you have written so answers can build off of that. To get started, this implementation parses the file and organizes by top level chain and residues. The counts are collapsed with addition for identical residues, so the output for your example looks like:

% python2.6 parse_dimplot.py dimplot-out.txt
{('C', '207', 'TRP'): {'M-M': 1, 'M-S': 0, 'Non-bonded contacts': 4, 'S-S': 0},
 ('C', '208', 'PHE'): {'M-M': 0, 'M-S': 0, 'Non-bonded contacts': 2, 'S-S': 0},
 ('L', '107', 'HIS'): {'M-M': 0, 'M-S': 0, 'Non-bonded contacts': 2, 'S-S': 0},
 ('L', '142', 'GLY'): {'M-M': 1, 'M-S': 0, 'Non-bonded contacts': 3, 'S-S': 0},
 ('L', '144', 'PRO'): {'M-M': 0, 'M-S': 0, 'Non-bonded contacts': 1, 'S-S': 0}}

And the code:

import sys
import pprint

def _parse_chain_parts(c):
    return c.replace("[", "").replace(" -> ", "").split()

def _parse_stats(s):
    m_m, s_s, m_s, non_bond = s.rstrip().split()
    return {"M-M": int(m_m), "S-S": int(s_s), "M-S": int(m_s),
            "Non-bonded contacts": int(non_bond)}

def parse_dimplot_line(line):
    line = line.replace("Residues: ", "")
    chain1, chain2, stats_str = line.split("]")
    stats = _parse_stats(stats_str)
    for chain_str in [chain1, chain2]:
        residue, r_num, chain = _parse_chain_parts(chain_str)
        yield (chain, r_num, residue, stats)

def parse_dimplot(in_file):
    with open(in_file) as in_handle:
        for line in in_handle:
            if line.startswith("Residues:"):
                for chain in parse_dimplot_line(line):
                    yield chain

def combine_stats(one, two):
    combined = dict()
    for k, val in one.items():
        combined[k] = val + two[k]
    return combined

in_file = sys.argv[1]
organized = dict()
for chain, r_num, residue, stats in parse_dimplot(in_file):
    r_key = (chain, r_num, residue)
    if organized.has_key(r_key):
        organized[r_key] = combine_stats(stats, organized[r_key])
    else:
        organized[r_key] = stats
pprint.pprint(organized)

You might decide to just use a standard library like BioPerl. If using Perl, you can create a mapping hash however you want and then proceed with your analysis using the code samples below.

my %map=(atom1=>[0,0,0,2], ...);

http://www.bioperl.org/Core/Latest/bptutorial.html#iii_9_1_using_3d_structure_objects_and_reading_pdb_files__structurei__structure__io_

  use Bio::Structure::IO;
  use strict;
  my $structio = Bio::Structure::IO->new(-file => $file);
  my $struc = $structio->next_structure;
  for my $chain ($struc->get_chains) {
     my $chainid = $chain->id;
     # one-letter chaincode if present, 'default' otherwise
     for my $res ($struc->get_residues($chain)) {
        my $resid = $res->id;
        # format is 3-lettercode - dash - residue number, e.g. PHE-20
        my $atoms = $struc->get_atoms($res);
        # actually a list of atom objects, used here to get a count
        print join "\t", $chainid,$resid,$atoms,"\n";
     }
  }

http://doc.bioperl.org/releases/bioperl-current/bioperl-live/Bio/Structure/IO/pdb.html

$stream = Bio::Structure::IO->new(-file => $filename,
                                  -format => 'PDB');

while (my $structure = $stream->next_structure) {
        # do something with $structure
}