Phred Quality Score Table
A Phred quality score converts a sequencing base-call error probability into a compact Q score: Q = -10 x log10(P_error). The table below shows the common FASTQ quality scores, their expected error rate, base-call accuracy and ASCII characters for Phred+33 and Phred+64 encodings.
| Q score | Error rate | P(error) | Accuracy | Phred+33 | Phred+64 |
|---|---|---|---|---|---|
| Q10 | 1 in 10 | 10% | 90% | + | J |
| Q20 | 1 in 100 | 1% | 99% | 5 | T |
| Q25 | 1 in 316 | 0.316% | 99.684% | : | Y |
| Q30 | 1 in 1,000 | 0.1% | 99.9% | ? | ^ |
| Q35 | 1 in 3,162 | 0.0316% | 99.968% | D | c |
| Q40 | 1 in 10,000 | 0.01% | 99.99% | I | h |
| Q50 | 1 in 100,000 | 0.001% | 99.999% | S | r |
| Q60 | 1 in 1,000,000 | 0.0001% | 99.9999% | ] | | |
Common quality bands
| Range | Label | Interpretation |
|---|---|---|
| Q10-Q19 | Low | Often trimmed or treated cautiously. |
| Q20-Q29 | Usable | Common minimum for rough variant or read filtering. |
| Q30-Q39 | High | A widely used benchmark for high-quality Illumina bases. |
| Q40+ | Very high | Very low expected base-call error rate. |
How FASTQ stores quality scores
A FASTQ record stores one quality character for every base. To decode a Phred+33 file, subtract 33 from each ASCII code; for Phred+64, subtract 64. Choosing the wrong encoding shifts every quality score by 31, which is why old Illumina FASTQ files need special care.
Frequently asked questions
What does a Phred quality score mean?
A Phred quality score Q is a logarithmic estimate of base-call error probability: Q = -10 x log10(P_error). Higher Q means a lower chance that the base call is wrong.
What does Q30 mean?
Q30 means the estimated error probability is 1 in 1,000, or 0.1%. Put another way, the expected base-call accuracy is 99.9%.
What is the difference between Phred+33 and Phred+64?
They are ASCII encodings for storing Q scores as characters in FASTQ files. Modern Sanger / Illumina 1.8+ FASTQ uses Phred+33. Older Illumina 1.3-1.7 data used Phred+64.
How do I know which FASTQ quality encoding I have?
Most current FASTQ files are Phred+33. If very low-quality characters such as !, " or # appear, the file must be Phred+33. Phred+64 is mainly relevant for old Illumina datasets.
Learn more
Sources
- 1Base-calling of automated sequencer traces using phred. II. Error probabilitiesEwing B, Green P · Genome Research 8(3):186-194; doi:10.1101/gr.8.3.186; PMID 9521922 · 1998The origin of the Phred quality score itself — the H1 formula 'Q = -10 x log10(P_error)' and FAQ item 1. This is the paper that defined a per-base-call error probability and the log-transformed quality value that every later FASTQ encoding stores. Cite it as the source of the definition, not of the ASCII table.
- 2The Sanger FASTQ file format for sequences with quality scores, and the Solexa/Illumina FASTQ variantsCock PJA, Fields CJ, Goto N, Heuer ML, Rice PM · Nucleic Acids Research 38(6):1767-1771; doi:10.1093/nar/gkp1137; PMID 20015970 · 2010The Phred+33 and Phred+64 columns of PHRED_ROWS in full (every character from '+'/'J' at Q10 to ']'/'|' at Q60), FAQ item 3, FAQ item 4, and the 'How FASTQ stores quality scores' section including the subtract-33 / subtract-64 decoding rule and the 31-position shift. Fetched text: "Sanger FASTQ files use ASCII 33-126 to encode PHRED qualities from 0 to 93 (i.e. PHRED scores with an ASCII offset of 33)" and "The Illumina 1.3+ FASTQ variant encodes PHRED scores with an ASCII offset of 64, and so can hold PHRED scores from 0 to 62 (ASCII 64-126)." It also states Q_PHRED = -10 log10 P_e, so it independently backs the H1 formula. Note both encodings' documented ranges contain every row on the page (Q60 is ASCII 93 in Phred+33 and ASCII 124 in Phred+64), and the Phred+64 range starting at ASCII 64 is what makes FAQ item 4's '!, " or # implies Phred+33' test valid.
- 3Quality Scores for Next-Generation Sequencing (Technical Note: Sequencing, Pub. No. 770-2011-030)Illumina, Inc. · 2011The Q10, Q20 and Q30 rows of PHRED_ROWS (error rate and accuracy columns), FAQ item 2, and the QUALITY_BANDS rows for Q20-Q29 and Q30-Q39. Its Table 1 'Quality Scores and Base Calling Accuracy' gives Q10 = 1 in 10 = 90%, Q20 = 1 in 100 = 99%, Q30 = 1 in 1,000 = 99.9%, alongside 'Q = - 10 log10 P'. It is also the primary source for the page's characterisation of Q30 as a benchmark: "When sequencing quality reaches Q30, virtually all of the reads will be perfect, having zero errors and ambiguities. This is why Q30 is considered a benchmark for quality in next-generation sequencing." Cite this as vendor documentation, which is what it is — it is the origin of the Q30 convention, not a peer-reviewed result. Verified by extracting the PDF text locally; document footer reads "Pub. No. 770-2011-030 Current as of 31 October 2011".
Related tools and references
Tools
Summarize and validate FASTA or FASTQ: counts, N50, GC, quality.
Per-base quality, GC and length distributions, duplication levels, overrepresented sequences and adapter content, each with a warn/fail verdict.
Inspect chromatograms, compare a read with its reference, or screen a whole plate of clones.