Why Science? | FAI and Random Access to FASTA Sequences

Science learning becomes powerful when students can tell a measured pattern from a model, a statistical decision from a mechanism, and a useful lead from a finished conclusion. An FAI file is a tab-delimited text index that enables efficient random access to well-formatted FASTA and FASTQ sequences. HTSlib documents five columns for FASTA—NAME, LENGTH, OFFSET, LINEBASES and LINEWIDTH—and a sixth QUALOFFSET for FASTQ. These fields let software calculate a byte position from a sequence coordinate and line layout. The index is tied to the exact source bytes: rewrapping lines, changing newlines, renaming records or rewriting the file can invalidate it even if the visible biological sequence seems unchanged. This guide supports Primary Science, PSLE Science, Secondary Science, O-Level Science and STEM education as a longform evidence-reading route; it is not medical advice, a laboratory protocol, an admission promise or a career guarantee.

Related eduKate reading: refget and globally identifiable reference sequences; BGZF and blocked compression for genomic data; SAM and human-readable sequence alignments; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.

Primary and official sources checked on 12 October 2026: Official HTSlib faidx file-format manual; Official samtools faidx command manual; Official SAM/BAM format specification for reference dictionary context; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Specifications, tools and recommendations can change, so preserve the dated documentation, exact source bytes, parser rules and software versions used.

A deeper reading route

FAI works because regularly wrapped text has predictable geometry. If each full sequence line contains the same number of bases and occupies a known number of bytes including its terminator, software can calculate how many complete lines precede a target position and how far into the next line to seek. It turns a large text file into something close to an addressable array without copying the sequences into the index.

NAME and LENGTH describe the logical sequence, while OFFSET, LINEBASES and LINEWIDTH describe its physical representation. OFFSET is the zero-based byte position of the first sequence base. LINEBASES counts bases per ordinary line; LINEWIDTH counts bytes, including newline bytes. Their difference commonly reveals LF versus CRLF but can also expose unusual terminators. These are byte facts, not genomic coordinates.

For FASTQ, QUALOFFSET adds the byte position of the first quality character. HTSlib requires quality lines to use the same wrapping as their corresponding sequence lines for indexing. Random access can then recover both a sequence interval and its quality substring. This capability does not solve ordinary paired-read lookup or validate the quality encoding.

The arithmetic is simple enough to audit. Convert the requested biological coordinate to the interface convention, calculate complete line blocks from LINEBASES, add newline bytes through LINEWIDTH and then add the within-line remainder to OFFSET. A single off-by-one, wrong newline width or coordinate convention can return letters that look believable because neighbouring sequence is also valid DNA.

Compression changes the address space. Plain FASTA uses ordinary byte offsets; bgzip-compressed FASTA can use additional .gzi information so a library maps uncompressed positions into BGZF blocks. The FAI and any compression index must match the exact object. Ordinary gzip cannot simply be treated as a seekable plain file.

Freshness is therefore a scientific integrity property. A stale .fai can keep the same filename and still point into the wrong record after rewrapping or edits. Robust workflows pair file and index hashes, rebuild after any rewrite, and compare indexed retrieval with a sequential parser at known boundaries.

Performance testing should distinguish index construction, cold-cache retrieval, warm-cache retrieval and complete scans. A tiny interval query may be dominated by storage latency, while a broad query can approach sequential reading cost. Remote storage adds range-request behaviour and object-version stability. Benchmarks should record interval width, file size, cache state, compression, hardware and returned sequence equality so speed remains interpretable rather than becoming a context-free headline.

Defensive implementations also bound every arithmetic operation. Corrupt lengths, offsets or line widths can produce overflow, seeks beyond the object or enormous allocations. A safe reader checks that fields are non-negative and internally possible, that OFFSET and QUALOFFSET lie within the source, and that the calculated final base does not exceed file size. Structural failure should be reported clearly instead of returning a plausible neighbouring substring.

Concurrency adds another practical layer. Several workers may safely read one immutable indexed reference, but replacing the FASTA or its index while queries are running can mix object generations. Atomic publication, immutable object versions and cache invalidation protect readers from pairing an old FAI with new bytes. A service should expose which object version answered a query so failures can be reproduced. That receipt should include the requested name, interval, source digest and index digest.

For students, FAI is a delightful coordinate puzzle. The index records where a chapter begins, how many letters fit on each printed line and how many bytes each line occupies. Mathematics then takes us directly to the requested character. The scientific lesson is equally important: an index is trustworthy only while it stays bound to the exact book edition.

A four-step claim ladder for this method

Stage	Question	Evidence	Responsible output
Describe	Which sequence and layout are indexed?	NAME, LENGTH, LINEBASES and LINEWIDTH	A logical and physical record model
Locate	Where do sequence or quality symbols begin?	OFFSET and optional QUALOFFSET	A byte-addressable start
Retrieve	Which bytes contain the requested interval?	Coordinate conversion and line arithmetic	An efficient subsequence result
Verify	Does the index still match the source?	Hashes, boundary tests and full-scan comparison	A trustworthy access path