Added ATCGN charset

  learning tasks.

  """

  def __init__(self, featurizer: Featurizer = OneHotFeaturizer, protein = False, max_length = 100):

  def __init__(self, featurizer: Featurizer = OneHotFeaturizer, charset: str = "ATCGN"):

    """Initialize FASTALoader.

    Parameters

      Whether or not the sequence passed in is a protein sequence. If False,

      it is treated as a nucleic acid sequence.

    max_length: int (default: 100)

      The maximum length of a string in the FASTA file. If Featurizer = OneHotFeaturizer,

      all one hot encodings will be padded to this length.

    charset: str (default: ATCGN)

      The charset used in the loaded FASTA file. Currently, we support ATCGN,

      full FASTA-format protein sequences, and full FASTA format nucleic acid

      sequences.

      Currently acceptable charsets are: "protein", "nucleic", and "ATCGN".

    """

    protein_charset = ('A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K',

                       'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V',

                       'W', 'X', 'Y', 'Z', '*', '-')

    nucleic_charset = ('A', 'C', 'G', 'T', 'U', '(i)', 'R', 'Y', 'K', 'M', 'S',

                       'W', 'B', 'D', 'H', 'V', 'N', '-')

    charsets = {

      "protein": ('A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L',

                  'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X',

                  'Y', 'Z', '*', '-'),

      "nucleic": ('A', 'C', 'G', 'T', 'U', '(i)', 'R', 'Y', 'K', 'M', 'S', 'W',

                  'B', 'D', 'H', 'V', 'N', '-'),

      "ATCGN": ('A', 'T', 'C', 'G', 'N')

    }

    self.charset = charsets.get(charset)

    max_length = len(self.charset)

    self.user_specified_features = None

    if isinstance(featurizer, UserDefinedFeaturizer):

      self.user_specified_features = featurizer.feature_fields

    if (featurizer == OneHotFeaturizer):

      if (protein):

        featurizer = OneHotFeaturizer(protein_charset, max_length)

      else:

        featurizer = OneHotFeaturizer(nucleic_charset, max_length)

      logger.warning(f"CHARSET IS {self.charset}")

      featurizer = OneHotFeaturizer(self.charset, max_length)

    else:

      featurizer = featurizer()

    self.featurizer = featurizer

  def create_dataset(self,

      sequences = np.array([])

      for input_file in input_files:

        sequences = np.append(sequences, _read_file(input_file))

      logger.warning(f"**TESTING** FEATURIZING ARRAY: {sequences}")

      X = self.featurizer(sequences)

      logger.warning(f"**TESTING** FINAL FEATURIZED ARRAY: {X}")

      logger.warning(X)

      ids = np.ones(len(X))

      # (X, y, w, ids)

      yield X, None, None, ids