SAM - Sequence Alignment/Map format, is a text-based, Tab-delimited format for storing sequence alignment data. BAM is the binary form of SAM (smaller size to save disk space). SAM uses 1-based indices while BAM uses 0-based indices.
Header section:
- Each header line starts with â@â followed by a 2-letter code.
- @HD: Header line with general info such as SAM format version and sorting order.
- @SQ: Reference sequence dictionary, with info such as sequence name & length.
- @RG: Read group info such as sequencing platform.
- @PG: Program used to generate the alignment.
Alignments section - 11 mandatory fields/columns.
| Col | Field | Description | Example |
|---|---|---|---|
| 1 | QNAME | Query template name | GAII05_0002:1:113:7822:3886#0 |
| 2 | FLAG | bitwise FLAG score | 4 |
| 3 | RNAME | Ref seq name | chr1 |
| 4 | POS | 1-based start cor | 11699950 |
| 5 | MAPQ | Mapping quality | 60 |
| 6 | CIGAR | CIGAR string | 39M1D12M |
| 7 | RNEXT | Ref name of mate/next read | chr1 |
| 8 | PNEXT | Pos of mate/next read | 11700332 |
| 9 | TLEN | (Paired) insert size | 433 |
| 10 | SEQ | read seq | AATGTAAAAC⊠|
| 11 | QUAL | Phred33 quality of seq | HHHAAB^^%CCC⊠|
| 12 | OPT | Optional tags | XA:i:2, MD:Z:0T34G15 |
FLAG score (column 2)
- FLAG score can be confusing, itâs an integer number storing dense information about the read.
- It answers 11 yes/no questions with 0/1, and string those 0 or 1 together gives you a number in binary, which is the FLAG score (often converted to decimal). The 11 questions are list in the table below.
- A nice explanation of the FLAG score by Dr. Quinlan can be found here.
- For example, if a read from a single end sequencing is unmapped, and passes quality check and not a PCR duplicate, basically if it answers no (with a 0) to all but one question in the table - question 3 âis the read itself unmappedâ, which means it has a 1 (yes) at third position of the binary number, then all 0 for other positions - a number of â00000000100â, which translate to a FLAG score of 4 in decimal system.
| base2 | base10 | Question | Pair end seq only |
|---|---|---|---|
| 00000000001 | 1 | Read comes from paired sequencing | N |
| 00000000010 | 2 | Read mapped in a proper pair | Y |
| 00000000100 | 4 | Read itself is unmapped | N |
| 00000001000 | 8 | Readâs mate is unmapped | Y |
| 00000010000 | 16 | Strand (0 forward, 1 reverse) | N |
| 00000100000 | 32 | Strand of mate | Y |
| 00001000000 | 64 | This is the first read in the pair | Y |
| 00010000000 | 128 | This is the second read in the pair | Y |
| 00100000000 | 256 | The alignment is not primary | N |
| 01000000000 | 512 | Read fails platform/vendor quality checks | N |
| 10000000000 | 1024 | Read is a PCR/optical duplicate | N |
Table adapted from Dr. Quilanâs slide.
CIGAR string
The string stores the operations needed to turn reference string to the experimentally sequenced read.
- M: Alignment match, could be a seq (character) match or mismatch(SNP).
- I: Insertion.
- D: Deletion.
- N: Split or spliced region
- S: Soft clipping (clipped sequences present in SEQ).
- H: Hard clipping (clipped sequences not present in SEQ).
- P: Padding.
- =: Seq match.
- X: Seq mismatch (SNP).
Example:
Reference: ACCTGTC--TACCTTACG Read: ACCT-TCCATACTTTATC Action: MMMMDMMIIMMMMMMMSS CIGAR: 4M 1D2M2I 7M 2S