nx_phred.nx

nx_phred.nx -- Phred quality score encode / decode helpers. license_tier: INDEPENDENT_REDERIVE genealogy_id: international-research-sources/sam-spec-v1.4.6 + ewing-green-1998-phred G2.0a of NISHI_GENOMICS_SUBSTRATE_ROADMAP.md. Bridges to FASTQ / SAM real-world data: every sequencer-emitted base carries a Phred quality score Q = -10 * log10(P_error), encoded in FASTQ as ASCII byte (Q + offset). This primitive ships ONLY the encode/decode layer; probability conversions (Q -> P_error in fixed-point) require log/exp primitives queued for G2.0b. Two encodings: Phred+33 (Sanger / Illumina 1.8+): the modern universal default - ASCII = Q + 33 - Q range [0..93] maps to ASCII [33..126] = '!' to '~' - Used by ALL contemporary sequencers + SAM/BAM Phred+64 (legacy Illumina 1.3 / 1.5 / 1.7): historical only - ASCII = Q + 64 - Q range [0..62] maps to ASCII [64..126] = '@' to '~' - Substrate supports for archived data + cross-compatibility Universal Q-value benchmarks: Q10 = 90% accuracy (10% error) Q20 = 99% accuracy (1% error) <- variant-calling minimum threshold Q30 = 99.9% accuracy (0.1% error) <- typical Illumina target Q40 = 99.99% accuracy <- typical PacBio HiFi Q60 = 99.9999% accuracy <- consensus / theoretical max API: nx_phred33_encode(q) -> i64 Q [0..93] -> ASCII byte; -1 out of range nx_phred33_decode(ascii) -> i64 ASCII [33..126] -> Q; -1 out of range nx_phred64_encode(q) -> i64 Q [0..62] -> ASCII byte; -1 out of range nx_phred64_decode(ascii) -> i64 ASCII [64..126] -> Q; -1 out of range nx_phred33_encode_string(qs, n, out_ascii) -> i64 bulk encode nx_phred33_decode_string(ascii, n, out_qs) -> i64 bulk decode What G2.0a does NOT do (deferred): - Q -> P_error fixed-point conversion -- needs nx_fixed_point + log10 (G2.0b) - Per-position quality aggregation (consensus quality scoring) -- G2.0c