<p align="center">
<img width="500" height="250" src="./perbase.png">
</p>
[](https://docs.rs/perbase)
[](https://crates.io/crates/perbase)
[](https://anaconda.org/bioconda/perbase)
A highly parallelized utility for analyzing metrics at a per-base level.
If a metric is missing, or performance is lacking. Please file a bug/feature ticket in issues.
## Why?
Why `perbase` when so many other tools are out there? `perbase` leverages Rust's concurrency system to automagically parallelize over your input regions. This leads to orders of magnitude faster runtimes that scale with the compute resources that you have available. Additionally, `perbase` aims to be more accurate than other tools. E.g.: `perbase` counts DELs toward depth, `bam-readcount` does not, `perbase` does not count REF_SKIPs toward depth, `sambamba` does.
## Installation
```bash
conda install -c bioconda perbase
# OR
cargo install perbase
# OR
brew install perbase
```
You can also download a binary from the [releases](https://github.com/sstadick/perbase/releases) page.
## Tools
### base-depth
The `base-depth` tool walks over every position in the BAM/CRAM file and calculates the depth, as well as the number of each nucleotide at the given position. Additionally, it counts the numbers of Ins/Dels at each position.
The output columns are as follows:
| REF | The reference sequence name |
| POS | The position on the reference sequence |
| REF_BASE | The reference base at the position, column excluded if no reference was supplied |
| DEPTH | The total depth at the position SUM(A, C, T, G, N, R, Y, S, W, K, M, DEL) |
| A | Total A nucleotides seen at this position |
| C | Total C nucleotides seen at this position |
| G | Total G nucleotides seen at this position |
| T | Total T nucleotides seen at this position |
| N | Total N nucleotides seen at this position |
| R | Total R nucleotides seen at this position |
| Y | Total Y nucleotides seen at this position |
| S | Total S nucleotides seen at this position |
| W | Total W nucleotides seen at this position |
| K | Total K nucleotides seen at this position |
| M | Total M nucleotides seen at this position |
| INS | Total insertions that start at the base to the right of this position |
| DEL | Total deletions covering this position |
| REF_SKIP | Total reference skip operations covering this position |
| FAIL | Total reads failing filters that covered this position (their bases were not counted toward depth) |
| NEAR_MAX_DEPTH | Flag to indicate if this position came within 1% of the max depth specified |
```bash
perbase base-depth ./test/test.bam
```
Example output
```text
REF POS DEPTH A C G T N R Y S W K M INS DEL REF_SKIP FAIL NEAR_MAX_DEPTH
chr1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 2 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 3 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 4 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 5 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 6 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 7 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 8 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
chr1 9 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 false
```
If the `--mate-fix` flag is passed, each position will first check if there are any mate overlaps and choose the mate with the hightest MAPQ, breaking ties by choosing the first mate that passes filters. Mates that are discarded are not counted toward `FAIL` or `DEPTH`.
If the `--reference-fasta` is supplied, the `REF_BASE` field will be filled in. The reference must be indexed and match the BAM/CRAM header of the input.
The output can be compressed and indexed as follows:
```bash
perbase base-depth -Z ./test/test.bam -o output.tsv.gz
tabix -S 1 -s 1 -b 2 -e 2 ./output.tsv.gz
# Query all positions overlapping region
tabix output.tsv.gz chr1:5-10
```
If the `--mate-fix` flag is passed, each position will first check if there are any mate overlaps and resolve based on the `mate-resolution-strategy`, mates that are discarded are not counted toward `FAIL` or `DEPTH`.
All strategies first check user-based read filters. If one mate fails filters, the other is chosen. If both fail, the first mate is chosen by default. For reads that are deletions / ref skips or lack a base call, all strategies fall back to the
Original strategy (MAPQ → first in pair).
| **BaseQualMapQualFirstInPair** | Higher base quality | Higher MAPQ | First mate in pair | Standard quality-first approach |
| **BaseQualMapQualIUPAC** | Higher base quality | Higher MAPQ | IUPAC code (e.g., A+G→R) | Returns ambiguity codes for ties |
| **BaseQualMapQualN** | Higher base quality | Higher MAPQ | N (unknown base) | Conservative, marks ambiguous as N |
| **MapQualBaseQualFirstInPair** | Higher MAPQ | Higher base quality | First mate in pair | Prioritizes mapping confidence |
| **MapQualBaseQualIUPAC** | Higher MAPQ | Higher base quality | IUPAC code (e.g., A+G→R) | Mapping-first with ambiguity codes |
| **MapQualBaseQualN** | Higher MAPQ | Higher base quality | N (unknown base) | Mapping-first, conservative |
| **IUPAC** | — | — | IUPAC code | Always returns IUPAC code for different bases, same bases return themselves (A+A→A) |
| **N** | — | — | N or base | Returns N for different bases, same bases return themselves (A+A→A) |
| **Original** | Higher MAPQ | First mate in pair | First mate (default) | Simple MAPQ-based strategy |
#### IUPAC Ambiguity Codes
When IUPAC strategies are used, the following codes are returned for base combinations:
| A | G | R | puRine (A or G) |
| C | T | Y | pYrimidine (C or T) |
| G | C | S | Strong (G or C) |
| A | T | W | Weak (A or T) |
| G | T | K | Keto (G or T) |
| A | C | M | aMino (A or C) |
| Any | Same | Original | Identical bases return themselves |
| Any | Other | N | Any combination not listed above |
#### Strategy Selection Guide
- **Use BaseQual strategies** when base quality is the most reliable indicator of accuracy
- **Use MapQual strategies** when mapping quality is more trustworthy (e.g., repetitive regions)
- **Use IUPAC variants** when you want to preserve ambiguity information for downstream analysis
- **Use N variants** when you prefer conservative base calling
- **Use FirstInPair variants** when you want deterministic results without ambiguity codes
- **Use IUPAC/N strategies** when you don't trust quality scores and want base-only decisions
- **Use Original** for backwards compatibility or simple MAPQ-based selection
#### Usage:
```text
perbase-base-depth 0.10.3
Seth Stadick <sstadick@gmail.com>
Calculate the depth at each base, per-nucleotide
USAGE:
perbase base-depth [FLAGS] [OPTIONS] <reads>
FLAGS:
-Z, --bgzip
Optionally bgzip the output
-h, --help
Prints help information
-k, --keep-zeros
Keep positions even if they have 0 depth
-m, --mate-fix
Fix overlapping mates counts, see docs for full details
-M, --skip-merging-intervals
Skip merging together regions specified in the optional BED or BCF/VCF files.
**NOTE** If this is set it could result in duplicate output entries for regions that overlap. **NOTE** This
may cause issues with downstream tooling.
-V, --version
Prints version information
-z, --zero-base
Output positions as 0-based instead of 1-based
OPTIONS:
-B, --bcf-file <bcf-file>
A BCF/VCF file containing positions of interest. If specified, only bases from the given positions will be
reported on
-b, --bed-file <bed-file>
A BED file containing regions of interest. If specified, only bases from the given regions will be reported
on
-C, --channel-size-modifier <channel-size-modifier>
The fraction of a gigabyte to allocate per thread for message passing, can be greater than 1.0 [default:
0.15]
-c, --chunksize <chunksize>
The ideal number of basepairs each worker receives. Total bp in memory at one time is (threads - 2) *
chunksize [default: 1000000]
-L, --compression-level <compression-level>
The level to use for compressing output (specified by --bgzip) [default: 2]
-T, --compression-threads <compression-threads>
The number of threads to use for compressing output (specified by --bgzip) [default: 4]
-F, --exclude-flags <exclude-flags>
SAM flags to exclude, recommended 3848 [default: 0]
-f, --include-flags <include-flags>
SAM flags to include [default: 0]
-M, --mate-resolution-strategy <mate-resolution-strategy>
If `mate_fix` is true, select the method to use for mate fixing [default: original]
-D, --max-depth <max-depth>
Set the max depth for a pileup. If a positions depth is within 1% of max-depth the `NEAR_MAX_DEPTH` output
field will be set to true and that position should be viewed as suspect [default: 100000]
-Q, --min-base-quality-score <min-base-quality-score>
Minium base quality for a base to be counted toward [A, C, T, G]. If the base is less than the specified
quality score it will instead be counted as an `N`. If nothing is set for this no cutoff will be applied
-q, --min-mapq <min-mapq>
Minimum MAPQ for a read to count toward depth [default: 0]
-o, --output <output>
Output path, defaults to stdout
--ref-cache-size <ref-cache-size>
Number of Reference Sequences to hold in memory at one time. Smaller will decrease mem usage [default: 10]
-r, --ref-fasta <ref-fasta>
Indexed reference fasta, set if using CRAM
-t, --threads <threads>
The number of threads to use [default: 10]
ARGS:
<reads>
Input indexed BAM/CRAM to analyze
```
### only-depth
The `only-depth` tool walks over the input BAM/CRAM file and calculates the depth over all positions specified by either a BED file or in the BAM/CRAM header. Adjacent positions that have the same depth will be merged together to form a non-inclusive range (see example output).
There are two distinct modes that `only-depth` can run in, gated by the `--fast-mode` flag. When running in fast-mode, only depth over the area a read covers is only determined by the reads start and end postions, and no cigar related info is taken into account. `--mate-fix` may still be used in this mode, and areas where mates overlap will not be counted twice.
Without the `--fast-mode` flag, the depth at each position is determined in a manner similar to `base-depth` where `DEL` will count toward depth, but `REF_SKIP` will not. Additionally, any reads that fail the `--exclude-flags` will not be counted toward depth. Lastly, `--mate-fix` can be applied to avoid counting regions twice where mates may overlap.
Regarding mate fixes, `perbase` will make "fixes" based only on the counted regions in a read. For example, if you have a read that goes from "chr1:0-1000" with a CIGAR of "25M974N1M", and the mate aligns nicely at "chr1:45-70" with CIGAR "25M", the mate will count toward the depth over "chr1:45-74". This is in contrast to other tools that will reject the mate even though it overlaps a region of R1 that is not counted toward depth.
For the fastest possible output, use `only-depth --fast-mode`.
**Note** that it is possible that two adjacent positions may not merge if they fall at a `--chunksize` boundary. If this is an issue you can set the `--chunksize` to the size of the largest contig in question. At a future date this may be fixed or a post processing tool may be provided to fix it. For most use cases this should not be a problem. Additionally, you can pipe into `merge-adjacent` which will fix it as well. EX: `perbase only-depth -m file.bam | perbase merge-adjacent > out.tsv`.
Example output of `perbase only-depth --mate-fix --zero-base ./test/test.bam`:
```text
REF POS END DEPTH
chr2 0 4 1
chr2 4 9 2
chr2 9 12 3
chr2 12 14 2
chr2 14 17 3
chr2 17 19 4
chr2 19 23 5
chr2 23 34 4
chr2 34 39 3
chr2 39 49 1
chr2 49 54 2
chr2 54 64 3
chr2 64 74 4
chr2 74 79 3
chr2 79 84 2
chr2 84 89 1
```
If a BED-like output is needed, `--bed-format -z` flags can be set, which will write a 0-based, no-header TSV output with an empty 4th column and the depth in the 5th column.
Usage:
```text
Calculate the only the depth at each base
USAGE:
perbase only-depth [FLAGS] [OPTIONS] <reads>
FLAGS:
--bed-format
Output BED-like output format with the depth in the 5th column. Note, `-z` can be used with this to change
coordinates to 0-based to be more BED-like
-Z, --bgzip
Optionally bgzip the output
-x, --fast-mode
Calculate depth based only on read starts/stops, see docs for full details
-h, --help
Prints help information
-k, --keep-zeros
Keep positions even if they have 0 depth
-m, --mate-fix
Fix overlapping mates counts, see docs for full details
-n, --no-merge
Skip merging adjacent bases that have the same depth
-M, --skip-merging-intervals
Skip mergeing togther regions specified in the optional BED or BCF/VCF files.
**NOTE** If this is set it could result in duplicate output entries for regions that overlap. **NOTE** This
may cause issues with downstream tooling.
-V, --version
Prints version information
-z, --zero-base
Output positions as 0-based instead of 1-based
OPTIONS:
-B, --bcf-file <bcf-file>
A BCF/VCF file containing positions of interest. If specified, only bases from the given positions will be
reported on. Note that it may be more efficient to calculate depth over regions if your positions are
clustered tightly together
-b, --bed-file <bed-file>
A BED file containing regions of interest. If specified, only bases from the given regions will be reported
on
-C, --channel-size-modifier <channel-size-modifier>
The fraction of a gigabyte to allocate per thread for message passing, can be greater than 1.0 [default:
0.001]
-c, --chunksize <chunksize>
The ideal number of basepairs each worker receives. Total bp in memory at one time is (threads - 2) *
chunksize [default: 1000000]
-L, --compression-level <compression-level>
The level to use for compressing output (specified by --bgzip) [default: 2]
-T, --compression-threads <compression-threads>
The number of threads to use for compressing output (specified by --bgzip) [default: 4]
-F, --exclude-flags <exclude-flags>
SAM flags to exclude, recommended 3848 [default: 0]
-f, --include-flags <include-flags>
SAM flags to include [default: 0]
-q, --min-mapq <min-mapq>
Minimum MAPQ for a read to count toward depth [default: 0]
-o, --output <output>
Output path, defaults to stdout
-r, --ref-fasta <ref-fasta>
Indexed reference fasta, set if using CRAM
-t, --threads <threads>
The number of threads to use [default: 32]
ARGS:
<reads>
Input indexed BAM/CRAM to analyze
```
## merge-adjacent
`merge-adjacent` is a utility to merge overlapping regions in a BED-like file.
It will take a file with four columns and no header as long as the columns are like:
```text
<contig>\t<start>\t<stop>\t<depth>\n
```
Or it can take files with three columns with headers that are like
```text
The `END|chromEnd` column is optional.
```text
perbase-merge-adjacent 0.7.5-alpha.0
Seth Stadick <sstadick@gmail.com>
Merge adjacent intervals that have the same depth. Input must be sorted like: `sort -k1,1 -k2,2n in.bed > in.sorted.bed`
Generally accepts any file with no header tha is <chrom>\t<start>\t<stop>\t<depth>. The <stop> is optional. See
documentation for explaination of headers that are accepted.
USAGE:
perbase merge-adjacent [FLAGS] [OPTIONS] [in-file]
FLAGS:
-Z, --bgzip
Optionally bgzip the output
-h, --help
Prints help information
-n, --no-header
Indicate if the input file does not have a header
-V, --version
Prints version information
OPTIONS:
-T, --compression-level <compression-level>
The level to use for compressing output (specified by --bgzip) [default: 2]
-T, --compression-threads <compression-threads>
The number of threads to use for compressing output (specified by --bgzip) [default: 32]
-o, --output <output>
The output location, defaults to STDOUT
ARGS:
<in-file>
Input bed-like file, defaults to STDIN
```
EX:
```bash
## Similar Projects
- [`sambamba depth`](https://github.com/biod/sambamba/wiki/%5Bsambamba-depth%5D-documentation)
- [`samtools depth`](http://www.htslib.org/doc/samtools-depth.html)
- [`mosdepth`](https://github.com/brentp/mosdepth)
- [`bam-readcount`](https://github.com/genome/bam-readcount)
