profile.build() — Build a profile for a given sequence or alignment

build(sdb, gap_penalties_1d=(-900.0, -50.0), matrix_offset=0.0, rr_file='$(LIB)/as1.sim.mat', n_prof_iterations=3, max_aln_evalue=0.1, matrix_scaling_factor=0.0069, check_profile=True, output_score_file=None, gaps_in_target=False, score_statistics=True, pssm_weights_type='HH1', pssm_file=None, window_size=1024)

This command iteratively scans a database of sequences to build a profile for the input sequence or alignment. The command calculates the score for a Smith-Waterman local alignment between the input sequence and each of the sequences in the database. The significance of the alignment scores (e-values) are calculated using a procedure similar to that described by [Pearson, 1998].

Alignments with e-values below max_aln_evalue are then added to the current alignment. A position-specific scoring matrix is then calculated for the current alignment and is used to search the sequence database. This procedure is repeated for n_prof_iterations or until there are are no significant alignments below the threshold, whichever occurs first.

The initial sequence or alignment can be read in either in the profile format, with profile.read(), or as an alignment using alignment.append(). In the latter case, the alignment has to be converted to the profile format using alignment.to_profile().

The output contains a multiple sequence alignment (assembled) of all the homologues of the input sequence found in the database. The output can be formatted as a profile with profile.write() or converted into any of the standard alignment formats using profile.to_alignment(). It can then be written out to a file with alignment.write().

The fit between the observed and theoretical distributions of the z-scores is calculated after each iteration and is reported in the log file. The fit is calculated using the Kolmogorov-Smirnov D-statistic. If the check_profile flag is set to True, then the command will not proceed if the fit deviates by more than 0.04 (D-statistic).

By default, regions of the alignment that introduce gaps in the target sequence are ignored (deleted) in the final multiple alignment. But if gaps_in_target is set to True, then the gaps are retained. (See below for comments).

The scores of each alignment between the input sequence and each database sequence, from all iterations, will be written out to the file specified in output_score_file (or if this is None, no such output will be written).

If pssm_file is set to a file name, the PSSM (position specific scoring matrix) calculated for each iteration is written to that file.

Comments:

1. The procedure has been optimized only for the BLOSUM62 similarity matrix.

2. The dynamic programming algorithm has been optimized for performance on Intel Itanium2 architecture. Nevertheless, the calculation is sufficiently CPU intensive. It takes about 20 min for an iteration, using an input sequence of 250aa against a database containing 500,000 sequences on an Itanium2 machine. It could take much longer on any other machine.

3. It is advisable to have gaps_in_target set to False when scanning against large databases, to avoid the local alignments inserting a huge number of gaps in the final alignments.

4. The statistics will not be accurate if the database does not have sequences that represent the entire range of lengths possible. In extreme cases, where statistics cannot be calculated at all, a StatisticsError will be raised.

5. The method can be used for fold-assignment by first building a profile for the target sequence by scanning against a large non-redundant sequence database (like swissprot) and then using the resulting profile to scan once against a database of sequences extracted from PDB structures. gaps_in_target can be set to True in the second step to get the complete alignments that can then be used for modeling.

See sequence_db.read() for a discussion of the window_size parameter.

Example: examples/commands/build_profile.py

from modeller import *
log.verbose()
env = environ()

#-- Prepare the input files

#-- Read in the sequence database
sdb = sequence_db(env)
sdb.read(seq_database_file='pdb95.fsa', seq_database_format='FASTA',
         chains_list='ALL', minmax_db_seq_len=(1, 40000), clean_sequences=True)

#-- Write the sequence database in binary form
sdb.write(seq_database_file='pdb95.bin', seq_database_format='BINARY',
          chains_list='ALL')

#-- Now, read in the binary database
sdb.read(seq_database_file='pdb95.bin', seq_database_format='BINARY',
         chains_list='ALL')

#-- Read in the target sequence/alignment
aln = alignment(env)
aln.append(file='toxin.ali', alignment_format='PIR', align_codes='ALL')

#-- Convert the input sequence/alignment into
#   profile format
prf = aln.to_profile()

#-- Scan sequence database to pick up homologous sequences
prf.build(sdb, matrix_offset=-450, rr_file='${LIB}/blosum62.sim.mat',
          gap_penalties_1d=(-500, -50), n_prof_iterations=5,
          check_profile=False, max_aln_evalue=0.01, gaps_in_target=False)

#-- Write out the profile
prf.write(file='buildprofile.prf', profile_format='TEXT')

#-- Convert the profile back to alignment format
aln = prf.to_alignment()

#-- Write out the alignment file
aln.write(file='buildprofile.ali', alignment_format='PIR')