How do you translate a SAM or BAM alignment report, explain a CIGAR string in another language and preserve exactly where a sequencing read aligns? Sequence Alignment/Map files connect individual read observations to reference sequences through machine-readable fields. Their names may look like English abbreviations and their rows may resemble ordinary text, but field positions, flags, coordinates and auxiliary tags have defined technical meanings. Translation should make those meanings accessible, not replace them with attractive text that changes the alignments.
Searches for SAM format translation, BAM file localization, CRAM reference decoding, SAM FLAG explained, CIGAR operations, MAPQ score, SAM tags, read groups, supplementary alignments and paired-end alignment interpretation often reflect different information needs. A learner wants a clear explanation. A laboratory wants a reproducible report. A developer wants localized interface labels without breaking parsers. A downstream analyst needs access to the same read coordinates and provenance that the original pipeline produced. An effective multilingual release serves each purpose without conflating documentation with data transformation.
This practical world-facing guide explains the SAM/BAM/CRAM ecosystem through reference dictionaries, QNAME, FLAG, RNAME, POS, MAPQ, CIGAR, RNEXT, PNEXT, TLEN, SEQ, QUAL, optional tags, sorting, indexing, supplementary alignments, reference-dependent compression and validation. It offers synthetic cases for training, but no real human sequencing data. Actual alignments must be interpreted with the producing mapper’s settings and the applicable GA4GH/HTS specification revision; evidence in an alignment file is not by itself a diagnosis.
The sixty-second answer
Translate the help text, captions, glossary and approved narrative. Leave alignment keys, sequence identifiers, CIGAR grammar, flag integers, reference names, coordinates, read-group links and binary record meaning unchanged. If format conversion, realignment, sorting, reference substitution or data repair is required, treat it as a distinct authorized workflow with reproducible validation and provenance.
1. Three representations and one family of alignment meaning
SAM is a human-readable text alignment representation, BAM is a compressed binary representation of alignment records, and CRAM is a reference-aware compressed representation with its own versioned specification. These are data formats, not three natural languages.
For example: An archive distributes one alignment dataset in BAM form and a translated tutorial showing the corresponding SAM fields for education.
The main mistake: A multilingual description that calls CRAM a translated SAM file confuses serialization, compression and linguistic localization.
Validate it by: Check which format and version the project actually stores, then compare parsed records rather than expecting identical bytes across formats.
2. SAM records have ordered fields
A SAM alignment line has eleven mandatory tab-separated fields followed by optional typed tags. The field order conveys meaning independently of the spoken language used in documentation.
In a practical setting: A Japanese guide labels columns with Japanese descriptions but displays original QNAME, FLAG, RNAME, POS, MAPQ, CIGAR, RNEXT, PNEXT, TLEN, SEQ and QUAL tokens.
The risk: Substituting translated headings or inserting prose columns into a SAM file prevents ordinary tools from decoding the intended records.
A robust check: Keep a canonical file unchanged; ensure the first eleven parsed fields retain their original positions and types.
3. Header lines begin with distinct @ records
SAM header lines such as @HD, @SQ, @RG and @PG describe sorting, reference sequences, read groups and program provenance. They are not generic annotations that may be freely rearranged.
Picture this case: A team adds French explanations of @SQ and @RG in a handbook, but keeps the underlying header type and tag keys stable.
An attractive but false shortcut: Replacing a header tag with a translated abbreviation can detach records from their reference or read-group definitions.
For sign-off: Parse the header and verify required links to references, read groups and programs before approving a localized release.
4. Reference sequence dictionaries are contracts
The @SQ header generally identifies reference sequence names and lengths, often alongside other metadata. RNAME and RNEXT values rely on those names, so consistency matters more than display language.
A reader might encounter: A chromosome called chr7 in the dictionary remains chr7 in records even when the documentation calls it chromosome seven in Spanish.
What must not happen: Renaming a sequence in one location only creates a file that appears readable but references a name its dictionary does not define.
To protect fidelity: Compare @SQ entries with the exact reference FASTA and confirm that every mapped RNAME resolves correctly.
5. A reference build shapes alignment meaning
The same textual chromosome label may exist in different genome assemblies with different sequences or positions. Read alignment is conditional on the chosen reference and mapping parameters.
For example: A data portal presents two assemblies separately and includes exact reference identifiers in each localized explanatory page.
The main mistake: Calling a set of alignments globally equivalent solely because both contain chr1 hides a potentially different coordinate system.
Validate it by: Record reference accession or build, check sequence dictionaries and hashes when available, and retain aligner provenance.
6. QNAME identifies a query template
The QNAME field identifies the query template or read-associated name under the format’s conventions. It is not an invitation to rewrite a person’s name or translate a sample label.
In a practical setting: A synthetic @readA FASTQ identifier can correspond to QNAME readA in a SAM explanation without changing the key.
The risk: Translating a substring inside QNAME can make it impossible to join the alignment to the original read or mate.
A robust check: Verify exact identifier relationships between FASTQ, alignments and any downstream records; do not use raw QNAMEs as public personal identifiers.
7. FLAG is a bitwise integer
FLAG encodes a combination of states in one integer. Different bits describe pairing, strand, mapped status, secondary alignment, QC failure, duplicate marking or supplementary alignment.
Picture this case: A French table explains that a flag may combine multiple states, rather than claiming that the number itself is a direct read-quality grade.
An attractive but false shortcut: Translating FLAG=99 into a free-text word without evaluating bits loses the combination of states and hides important distinctions.
For sign-off: Decode the integer using a standards-aware tool and test each relevant bit against the expected alignment and mate fields.
8. Paired-read flags are not the same as direction
Paired-template status, proper-pair status and mate mapping status are distinct concepts. Their definitions depend on record flags and the producing aligner.
A reader might encounter: A guide explains read-paired and properly-aligned-pair as separate indicators, without assuming that every pair is acceptable for a specific assay.
What must not happen: A translated phrase ‘correct pair’ can overpromise that the alignment is experimentally or clinically valid because a bit is set.
To protect fidelity: Use the mapper’s proper-pair policy, inspect both mates and keep the original flags in any exported data.
9. Unmapped and mapped statuses need careful language
FLAG bits distinguish unmapped reads and unmapped mates. Other fields can have special values or potentially meaningful coordinates even when a read is flagged unmapped.
For example: An educational example labels a read as unmapped while explaining why a nonempty reference-related field may still appear in its record.
The main mistake: Calling any nonzero POS a confirmed alignment ignores FLAG and the specification’s handling of unmapped records.
Validate it by: Interpret mapped status using FLAG first, then verify consistency of RNAME, POS, CIGAR and mate information.
10. Secondary versus supplementary alignment
Secondary alignments represent alternative mappings in one sense, while supplementary records represent segments of a chimeric or split alignment. The concepts should not be translated as synonyms.
In a practical setting: A report distinguishes a multi-mapped read from a split read that spans two alignment segments, using separate captions.
The risk: Replacing both labels with ‘extra alignment’ loses whether a record is an alternative mapping or part of a split structure.
A robust check: Check 0x100 and 0x800 status as appropriate, and validate related SA tags and primary alignment relationships.
11. RNAME names the reference target
RNAME identifies the reference sequence to which a segment is aligned, or an appropriate missing value where applicable. It is a computational reference key, not a descriptive place name.
Picture this case: A Spanish legend may say ‘reference chromosome’ but retain the exact RNAME token and sequence identifier.
An attractive but false shortcut: Renaming a contig in a translated export without updating and verifying the reference dictionary makes alignments incoherent.
For sign-off: Resolve every mapped RNAME against @SQ and the reference used by the pipeline.
12. POS is a one-based leftmost coordinate
For mapped alignments, POS identifies a one-based leftmost alignment coordinate on the reference. This is not necessarily the read’s original first base or a zero-based array index.
A reader might encounter: A read aligned at POS 100 starts at the hundredth reference base in SAM coordinates, subject to its CIGAR alignment operations.
What must not happen: Translating POS as zero-based position introduces off-by-one errors when coordinates are copied into programming libraries.
To protect fidelity: Test known positions across a reference FASTA and state any coordinate conversion convention explicitly.
13. MAPQ is alignment mapping quality
MAPQ represents mapping-quality information with Phred-scale interpretation under the format and mapper conventions. A special value can indicate unavailable mapping quality rather than perfect certainty.
For example: A chart can explain that high mapping quality may suggest less ambiguity while still reporting mapper-specific limitations.
The main mistake: Reading MAPQ 255 as ‘255 percent confidence’ is incorrect, and assuming all aligners calibrate MAPQ identically is unsafe.
Validate it by: Consult the aligner’s documentation and preserve raw MAPQ before displaying any human-readable description.
14. CIGAR describes how an alignment consumes sequence
The CIGAR string encodes runs of alignment operations such as M, I, D, N, S, H, P, = and X. Operation counts and ordering represent how query and reference positions interact.
In a practical setting: A translated tutorial explains 5M1I4M as five aligned positions, one query insertion and four more aligned positions under CIGAR semantics.
The risk: Rewriting CIGAR letters as translated words inside the file destroys the parseable operation sequence and changes position calculations.
A robust check: Use a CIGAR-aware library to compute query and reference consumption, then check it against SEQ and mapping coordinates.
15. M, equals and X must not be collapsed casually
M denotes alignment positions without necessarily distinguishing match and mismatch. The = and X operators explicitly represent matches and mismatches where used.
Picture this case: An explanation may clarify that 10M does not assert ten exact nucleotide matches, while 10= carries different detail.
An attractive but false shortcut: A translated sentence saying every M is a matching base overstates the available evidence and may conceal mismatches.
For sign-off: Interpret MD or reference-aware sequence comparison when exact base matching is required, keeping the mapper’s encoding choices visible.
16. Insertions and deletions consume different axes
CIGAR I consumes query sequence without consuming reference bases, while D consumes reference positions without consuming query bases. N indicates a skipped reference region.
A reader might encounter: A read with a one-base insertion can have a longer query-consuming CIGAR length than its reference-consuming span.
What must not happen: Translating both I and D as ‘missing base’ without distinguishing which sequence is affected reverses the event.
To protect fidelity: Compute query and reference spans separately and verify with a small illustrative alignment.
17. N commonly marks skipped reference regions
The N operator consumes reference but not query sequence and can be used for spliced alignments. Its interpretation depends on assay and aligner context.
For example: An RNA sequencing explanation shows two exonic alignment blocks separated by a reference skip, without claiming that the skipped bases were deleted from the genome.
The main mistake: Calling every N operator a deletion conflates splicing-related alignment representation with an actual sequence variant.
Validate it by: Check assay type and aligner behavior; distinguish skipped reference ranges from deletion operations in the CIGAR.
18. Soft and hard clipping preserve different information
S indicates soft-clipped query bases that appear in SEQ, whereas H indicates hard-clipped bases omitted from SEQ in the alignment record. Both influence how the visible sequence relates to the query.
In a practical setting: A guide shows a read with 5S95M and explains why its first five bases may be present in SEQ but not aligned.
The risk: Describing a clipped read as having permanently lost five biological bases may be misleading, especially when the original FASTQ remains available.
A robust check: Compare SEQ length with query-consuming CIGAR operations and inspect the original read where available.
19. Padded alignments use a special operation
The P operator represents padding in certain multiple-alignment representations. It is uncommon in many workflows but remains part of the formal CIGAR vocabulary.
Picture this case: A format glossary includes P as a defined operation rather than ignoring it because the team’s usual files do not contain it.
An attractive but false shortcut: An overzealous normalizer may discard a rare operator and silently change a special-case alignment representation.
For sign-off: Include rare but valid operators in parser tests and document whether recipient software supports them.
20. RNEXT and PNEXT describe the mate
RNEXT and PNEXT communicate reference identity and position for the next read in a template under the format’s conventions. They are not a second copy of the current segment’s location.
A reader might encounter: A bilingual training example labels the current read’s POS and its mate’s PNEXT separately, making the distinction visible.
What must not happen: Copying POS into PNEXT to make columns look consistent can break paired-end interpretation and downstream checks.
To protect fidelity: Validate both mate records and compare paired flags, RNEXT, PNEXT and alignment coordinates.
21. TLEN has a defined template meaning
TLEN gives an inferred template length with sign conventions that may matter in paired data. It is not simply the length of a sequenced read, nor always the same as a DNA insert in every context.
For example: A report explains template span from mapped outer coordinates and keeps separate the single read length of each mate.
The main mistake: Translating TLEN into ‘read length’ can reverse conclusions about fragment sizes or alignment geometry.
Validate it by: Compare mapped mate locations and the specification’s TLEN convention before explaining derived insert-size metrics.
22. SEQ contains the read sequence representation
SEQ holds base calls in the alignment record’s orientation and may use special omission conventions. It is biological data, not a translatable acronym.
In a practical setting: A teaching table displays an aligned SEQ string and its original FASTQ read to illustrate orientation and clipping.
The risk: Replacing SEQ letters with translated nucleotide names or assuming they are always the original read orientation changes meaning.
A robust check: Preserve SEQ exactly and consult orientation flags and original reads when interpreting biological direction.
23. QUAL values align with SEQ
QUAL provides per-base quality values as defined by the SAM representation and must correspond to the displayed sequence where present. A missing quality value is different from a low quality.
Picture this case: An editor explains that a read can have no available quality string in a particular exported record without inventing quality zero.
An attractive but false shortcut: Changing quality characters by typography or misaligning them after clipping distorts evidence attached to individual base calls.
For sign-off: Check SEQ/QUAL lengths, absence markers and parsed numeric quality values through a trusted tool.
24. Optional fields have TAG:TYPE:VALUE structure
Auxiliary fields convey typed annotations with two-character tags and value types. Some tags are standardized; others are program-specific and require external documentation.
A reader might encounter: A localizer explains NM or RG in another language while retaining the literal tag identifiers and serialized type markers.
What must not happen: A find-and-replace operation that changes a tag key or type character may make downstream algorithms interpret the value incorrectly.
To protect fidelity: Inventory the optional tags present and distinguish standard tags from private or pipeline-specific fields.
25. NM and MD provide different alignment clues
NM commonly counts edit distance under defined conventions, while MD is an alignment difference string that allows some reference mismatch information to be reconstructed. Neither is a free-form error message.
For example: A bilingual summary labels NM as edit-distance related rather than as a sample error code, and explains the need for reference context.
The main mistake: Translating MD letters or numbers as ordinary prose is likely to destroy the encoded mismatch pattern.
Validate it by: Compare auxiliary values with CIGAR, SEQ and the reference using appropriate validation tools.
26. Read groups support provenance and analysis
The RG tag connects an alignment to read-group metadata often used for libraries, samples or sequencing units. Read group IDs are linkage keys, not display-language phrases.
In a practical setting: A multilingual report says which library a read group represents and keeps the ID unchanged for data joins.
The risk: Reassigning RG identifiers while translating sample names can merge distinct libraries or detach reads from header records.
A robust check: Verify every RG value points to the correct @RG header record and corresponding manifest entry.
27. Program groups record processing history
The @PG header can document tools, versions and command chains that produced an alignment file. These records are technical provenance rather than narrative acknowledgements.
Picture this case: A Chinese-language methods section explains which mapper and sorter were used, quoting exact tool names and versions.
An attractive but false shortcut: A translated report that replaces software identifiers with localized generic descriptions can make reproducing the analysis impossible.
For sign-off: Preserve exact tool versions and parameters and cross-reference run logs rather than paraphrasing away the computational history.
28. SA tags describe supplementary relationships
The SA optional tag can point to other alignment segments in split or chimeric mappings. Its structured values may contain contig names, positions, strands and CIGAR strings.
A reader might encounter: A technical guide explains split-read evidence around a rearrangement using a separate diagram and unchanged SA records.
What must not happen: Treating commas and semicolons in SA values as prose punctuation can break parser recognition or split-partner interpretation.
To protect fidelity: Parse SA with suitable tools and reconcile its reported partners with supplementary records.
29. Duplicate marking does not necessarily delete reads
A duplicate-related FLAG bit can annotate reads according to a pipeline’s duplicate-detection policy. Marking and removing duplicates are separate operations.
For example: A quality report explains that duplicate-marked reads remain in the BAM file unless a later filtering step removes them.
The main mistake: A translation that says ‘duplicate reads were removed’ merely because a marking tool ran misrepresents data retention.
Validate it by: Inspect the pipeline commands and compare record counts before and after any actual filtering stage.
30. Quality-control failure bits need context
A QC-fail bit communicates a recorded quality-control state, not necessarily a universal verdict about all potential analyses. Criteria depend on the workflow.
In a practical setting: An international collaboration documents why a class of reads failed the instrument or analysis filter without converting the flag to a clinical judgment.
The risk: Replacing ‘failed vendor filter’ with ‘scientifically worthless’ removes important nuance and may overstate the source evidence.
A robust check: Capture filter provenance, retain literal flags, and distinguish instrument quality control from research interpretation.
31. Sorted order is metadata with operational effects
SAM/BAM files can be sorted by coordinate or query name, among other arrangements. Header declarations and actual order must agree for downstream operations.
Picture this case: A portal offers a coordinate-sorted BAM and names it accordingly in each language; its labels describe actual state rather than aspiration.
An attractive but false shortcut: Changing the sort-order text in a translated description does not sort records, and changing only a header claim can mislead tools.
For sign-off: Verify order through a library or validation command and record the transformation if sorting is performed.
32. BAM is binary and BGZF-compatible
BAM stores alignment records in a compressed binary form designed for efficient tool processing and indexing. It should not be treated as plain multilingual prose.
A reader might encounter: A team supplies a BAM download with Spanish documentation and a separately generated human-readable SAM excerpt.
What must not happen: Editing the compressed bytes in a text editor or renaming a .bam file to .sam does not constitute a valid conversion.
To protect fidelity: Use samtools or equivalent format-aware software to inspect, convert and compare record fields and headers.
33. Indexes enable regional retrieval
BAI and CSI indexes support random access to suitable alignment files. The appropriate index depends on reference lengths, format and software support.
For example: An online genome browser receives an alignment file and its matching index, both documented in German for local readers.
The main mistake: A stale index after file rewriting may return incomplete regions even if streaming the file still seems to work.
Validate it by: Regenerate or validate an index after any permitted rewrite and compare multiple reference intervals against full-file iteration.
34. CRAM may depend on the exact reference
CRAM achieves compression partly through representations that can use a reference sequence. Decoding may require matching reference data or embedded information, depending on the file.
In a practical setting: An archive ships a CRAM file with explicit reference accession and retrieval instructions in multiple languages.
The risk: Translating a reference identifier or supplying a near-match assembly can prevent decoding or produce incorrect interpretation.
A robust check: Test decoding with the exact required reference, record version and reference fingerprints where available.
35. Compression choices and quality preservation
CRAM can support lossless representation under suitable settings, and pipelines may optionally apply lossy transformations such as quality binning. Compression alone does not establish whether every original field survives.
Picture this case: A release note describes whether per-base qualities were preserved or deliberately simplified before a CRAM export.
An attractive but false shortcut: Calling all CRAM files lossy is inaccurate, but claiming full fidelity without checking encode settings is equally risky.
For sign-off: Compare decoded sequences and qualities with the source and document any intentional modifications during conversion.
36. Coordinate systems differ across tools
SAM alignment POS is one-based, while some APIs and BED intervals use zero-based half-open coordinates. A correct language translation cannot repair an unannounced coordinate mismatch.
A reader might encounter: A report converts an interval for a browser and labels both original SAM coordinates and target API convention.
What must not happen: A silent off-by-one conversion can change a splice junction or structural-breakpoint interpretation despite fluent explanatory prose.
To protect fidelity: Use synthetic control intervals and explicit conversion rules, then test boundary cases.
37. Read orientation is a flag-and-sequence question
Reverse-complement alignment presentation involves orientation flags and sequence conventions. A read’s direction in a reference browser should not be inferred from its name or file suffix.
For example: A figure describes why a segment aligns to the reverse strand while retaining its FLAG and observed alignment data.
The main mistake: Calling R2 a reverse-strand read in all experiments conflates mate number with mapped orientation.
Validate it by: Decode strand flags and compare with reference-aware read representation, using an example that includes both orientations.
38. Privacy still applies to alignment data
Human-derived alignment files can contain identifiable genetic information, sample names and processing metadata. Being compressed or binary does not make them safe to upload to public translation tools.
In a practical setting: An institution translates a public protocol and uses synthetic alignments for training while keeping real patient-derived BAM files inside authorized systems.
The risk: A developer may expose raw QNAMEs or tags in an otherwise anonymized multilingual screenshot.
A robust check: Inspect sample and header metadata, apply approved de-identification and access controls, and audit exported examples.
39. Round-trip comparison is stronger than a visual diff
A multilingual release must preserve semantic fields, references, read-group mappings and counts. A visual inspection of a few lines cannot detect every missing or shifted alignment.
Picture this case: A data steward converts BAM to SAM for review then checks the relevant parsed fields and returns a validated BAM copy when necessary.
An attractive but false shortcut: A different binary checksum after a codec conversion may reflect packaging rather than changed alignment records; the opposite can also hide a bad manifest.
For sign-off: Compare normalized record identities, CIGAR, flags, coordinates, tags and sequence/quality values with format-aware tooling.
40. A small CIGAR case teaches the central distinction
Consider a synthetic record aligning with 4M1I3M. It consumes eight query bases and seven reference bases, assuming the basic interpretation of those operators. The operation sequence is a miniature model of alignment semantics.
A reader might encounter: An instructor explains the mismatch between lengths in Portuguese, using a diagram rather than rewriting the literal CIGAR code.
What must not happen: A reader who assumes 4M1I3M always spans eight reference bases will misplace downstream reference coordinates.
To protect fidelity: Ask students to calculate both axis lengths, then verify their result with a CIGAR parser.
41. Multilingual software needs a protected-token policy
A translator-facing resource should mark SAM field names, CIGAR operations, flags, reference IDs and tag names as fixed, while allowing local-language descriptions and context-sensitive scientific terminology.
For example: A glossary presents ‘mapping quality’ in several languages with the protected token MAPQ and a caution about mapper-specific meaning.
The main mistake: A generic machine-translation pipeline run directly over .sam lines can alter punctuation, spacing and identifier strings.
Validate it by: Export only the approved user-interface strings and explanatory documentation; enforce validation around any data-adjacent serialization.
42. A trustworthy release has three reviewers
Technical accuracy, linguistic readability and data integrity are related but not identical. A correct workflow assigns each to people or tests equipped to assess it.
In a practical setting: A bilingual manual is reviewed by a native-language editor, a sequencing analyst and a data engineer who validates immutable alignments.
The risk: One reviewer skilled only in prose may miss a reversed alignment, while a parser alone cannot judge a confusing explanation.
A robust check: Document review sign-off, preserve source artifacts and version the translated help separately from the scientific data.
43. Know the boundary to downstream variant calls
SAM/BAM/CRAM store read-to-reference alignment evidence. VCF reports variants inferred through a later calling pipeline. One does not automatically substitute for the other.
Picture this case: A student traces a hypothetical FASTQ read through mapping to BAM and onward to a VCF variant call in a clearly labelled example.
An attractive but false shortcut: Saying that a mapped read is a confirmed mutation erases inference, calling and filtering stages.
For sign-off: Keep file-format roles distinct and attach each scientific claim to the processing stage that supports it.
44. Design a final verification exercise
A useful training exercise asks readers to decode a FLAG, compute CIGAR spans, identify mate position, interpret a missing quality field and find the reference build.
A reader might encounter: Students complete the same evidence-preserving questions in English and Spanish, with answer keys using unchanged field values.
What must not happen: An exercise that rewards natural wording but never tests semantic fidelity teaches an incomplete translation skill.
To protect fidelity: Evaluate exact field mapping, coordinate arithmetic and clarity together; require corrections when any one fails.
Six worked cases
A paired alignment with conflicting labels
A dataset has a properly paired record according to its mapper and a mate on the same reference. The multilingual summary must avoid treating the proper-pair flag as a universal biological guarantee. Examine both mates, CIGAR spans, RNEXT, PNEXT and TLEN; the reader should be able to reconstruct the actual relationship without changing a record.
A spliced RNA read
An RNA-seq alignment uses CIGAR N to skip a reference region between aligned blocks. In translation, describe the skip in the context of splicing evidence and the assay, not as a genomic deletion. Verify with the aligner and transcript context, and leave the CIGAR code untouched.
Supplementary evidence for a breakpoint
A read has primary and supplementary records connected through SA metadata. A localized diagram can describe split alignment evidence; it cannot infer a fully validated structural variant without additional calling and review. Reconcile record links and orientation before publishing the explanation.
A CRAM archive without its reference
A laboratory sends a CRAM file and a translated README that omits its reference. The recipient cannot reliably decode or interpret it under the expected workflow. Recover the exact reference accession and any required files, test decoding and then add clear multilingual retrieval instructions; do not substitute a guessed genome build.
A misleading MAPQ chart
An editor turns a mapper-specific MAPQ score into a percentage of clinical certainty. The graph sounds persuasive but changes the meaning. The corrected caption names mapping quality and its algorithmic context, and explicitly avoids disease or diagnostic inference.
A converted but not identical binary file
A BAM is re-encoded with different compression and receives a new checksum. This is not proof that the alignment semantics changed; compare parsed fields, header relationships and representative intervals. Label the conversion, keep both checksums and document that this is a format transformation rather than language translation.
Questions and answers
Can I translate the SAM field names?
Use translated labels in documentation and user interfaces, but keep the canonical machine-readable field positions and tokens in the data.
Is MAPQ a percentage?
No. It is a mapping-quality measure whose interpretation depends on format conventions and the producing aligner.
Does 10M mean ten matching bases?
Not necessarily. M does not itself distinguish exact matches from mismatches; see =, X and other evidence.
What distinguishes BAM from CRAM?
Both are alignment containers, but their binary storage and compression approaches differ; CRAM can rely on a matching reference for decoding.
Can a marked duplicate still exist in BAM?
Yes. Marking duplicates and removing them are different operations.
What is the first check after translation?
Verify that alignments, flags, coordinates, CIGAR, read-group links and the reference dictionary remain unchanged.
Why must the reference build be recorded?
Coordinates and alignment meaning depend on the exact reference; chromosome names alone are insufficient.
Should private alignment data go into a public translator?
Not without appropriate authorization. Use synthetic examples for public lessons and approved secure workflows for restricted data.
A practical multilingual alignment checklist
Lock the canonical data, file version and reference sequence dictionary; inventory all SAM headers and used optional tags; protect FLAG integers, coordinates, CIGAR strings, sequences and quality characters; translate explanations in a separate layer; compare record counts and read-group linkage; test reverse-strand, paired, clipped, spliced, secondary and supplementary cases; verify BAM/CRAM conversion and index validity; document any intentional transformation; and obtain technical, linguistic and privacy review.
Authoritative specifications and related reading
GA4GH / HTS: SAM, BAM, CRAM, tags and indexes
HTSlib documentation and utilities