Describe the bug

clean_protein_seq() in pyaptamer/utils/_pseaac_utils.py replaces invalid amino acid characters with "N". However, "N" is the IUPAC one-letter code for Asparagine a real amino acid already present in the AMINO_ACIDS list.

This silently corrupts protein sequences by turning unknown characters into a valid amino acid, producing incorrect PSeAAC feature vectors downstream for both PSeAAC and AptaNetPSeAAC.

To Reproduce

from pyaptamer.utils._pseaac_utils import clean_protein_seq, AMINO_ACIDS

# "N" is Asparagine a valid amino acid
print("N" in AMINO_ACIDS)  # True

# Input sequence with invalid character "X"
result = clean_protein_seq("ACXD")
print(result)  # "ACND" — X was silently replaced with Asparagine!

# This corrupts PSeAAC feature vectors:
from pyaptamer.pseaac import PSeAAC
p = PSeAAC()
# These produce DIFFERENT vectors even though they should be equivalent
vec_buggy = p.transform("ACNDACNDACNDACNDACND")  # contains false Asparagine
vec_clean = p.transform("ACDACDACDACDACDACDAC")  # without corruption

Expected behavior

Invalid characters should be removed (filtered out) from the sequence rather than replaced with a valid amino acid. The existing UserWarning should be kept to inform the user about the removal.

Note: the aa_str_to_letter() utility in the same package already correctly uses "X" for unknown amino acid codes, showing an inconsistency in how unknowns are handled.

Additional context

• AMINO_ACIDS = list("ACDEFGHIKLMNPQRSTVWY") N is Asparagine at position 11
• The bug affects all PSeAAC feature computations when input contains non-standard residues (e.g., X, B, Z, digits, whitespace)
• All 144 existing tests pass after applying the fix

Versions

pyaptamer 0.1.0a1