tfio.genome.phred_sequences_to_probability

Converts raw phred quality scores into base-calling error probabilities.

@tf.function
tfio.genome.phred_sequences_to_probability(
    phred_qualities
)

Used in the notebooks

Used in the tutorials

For each ASCII encoded phred quality score (X), the probability that there was an error calling that base is computed by:

P = 10 ^ (-(X - 33) / 10)

This is assuming an "ASCII base" of 33.

The input is a tf.string tensor of ASCII encoded phred qualities, one string per DNA sequence, with each character representing the quality of a nucelotide.

For example:

phred_qualities = [["BB<"], ["BBBB"]]

phred_qualities A tf.string tensor where each string represents the phred quality of a DNA sequence. Each character in the string is the ASCII representation of the phred quality number.

tf.RaggedTensor The quality scores for each base in each sequence provided.