SAMBAMBA-MARKDUP(1)

💅 SAMBAMBA-MARKDUP — the tea ☕

okay bestie, welcome to sambamba-markdup, a lil legend from the sambamba fam of manual pages. she's got ONE job and she does it gorgeously: finding duplicate reads in a BAM file and dealing with them. no cap. 🧚

✨ SYNOPSIS ✨

sambamba markdup OPTIONS <input.bam> <output.bam>

🌸 DESCRIPTION 🌸

so basically she marks duplicate reads — that's the default, she's polite like that — or, if you ask her nicely, she straight up removes them. gone. deleted. blocked. 🚫

and how does she decide if a read is a duplicate or not?? the same "sum of base qualities" method that Picard uses. iconic behaviour, honestly. you can read all about the method at the Picard metric definitions page:

https://broadinstitute.github.io/picard/picard-metric-definitions.html

👑 OPTIONS 👑

• -r, --remove-duplicates — ok so instead of just marking the duplicates, she actually removes them. like, physically. from the output. the audacity, the power 💀
• -t, --nthreads=NTHREADS — how many threads you wanna give her to play with. she'll use 'em. all of 'em.
• -l, --compression-level=N — set the compression level of the resulting file, and it's gotta be from 0 to 9. pick your vibe. ✨
• -p, --show-progress — shows you a lil progress bar in STDERR so you can watch her work and feel something 🫶
• --tmpdir=TMPDIR — tell her where to stash her temporary files. default is /tmp, which is fine, but if you've got opinions, this is where you share them.
• --hash-table-size=HASHTABLESIZE — the size of the hash table she uses for finding read pairs. default is 262144 reads. heads up: it gets rounded down to the nearest power of two, so don't get cute with weird numbers. for good performance it should be > (average coverage) * (insert size). do the math bestie 🔥
• --overflow-list-size=OVERFLOWLISTSIZE — this is the overflow list, aka the second-chance lounge 💖 where reads that got thrown out of the hash table get another opportunity to meet their pairs. default is 200000 reads. bumping the size up means she creates fewer temporary files. slay.
• --io-buffer-size=BUFFERSIZE — controls the sizes of two buffers, each one BUFFERSIZE megabytes, used for reading and writing BAM during the second pass. default is 128. yes, two. she's generous. 🌈

🦋 SEE ALSO 🦋

Picard, and specifically the metric definitions for removing duplicates:

https://broadinstitute.github.io/picard/picard-metric-definitions.html

go read it, it's giving educational. 📚

💀 BUGS 💀

bestie. listen. external sort is not implemented. so memory consumption grows by 2Gb per each 100M reads. that is not a drill. check that you actually have enough RAM before you run this tool, or your machine will simply pass away. 😭

ok that's the whole page, no notes. this one's from sambamba-markdup(1), the SAMBAMBA-MARKDUP(1) manual — dated February 2015 🎀 stay hydrated, check your RAM, and go mark some dups. periodt. 💅✨