3SEQtools

Pipelines for processing and QC of data from Smart-3SEQ or classic 3SEQ. Most scripts will show their usage options if you invoke them with the --help flag. Most scripts have hardcoded constants defined at the top so you can customize them to your needs.

FASTQ generation pipeline

make_fastq.sh

Wrapper to call bcl2fastq with preferred settings and simpler usage. Mostly used with NextSeq data so some features might not make sense on other platforms (e.g. --no-lane-splitting could be bad on HiSeq).

FASTQ alteration pipeline

umi_homopolymer.py

Removes the poly(A) tail, moves the UMI to the read metadata, and discards the G-overhang in every read of FASTQ data. Generates a simple log of trimmed read lengths. align_smart-seq.sh etc. invoke this but let the aligner do the poly(A) removal instead. Much faster with PyPy.

Alignment pipelines

align_smart-3seq.sh

Starts from a list of single-end Illumina FASTQ files and aligns them into sorted, indexed, duplicate-marked BAM files. samtools and STAR are expected to be in your PATH and UMI-dedup locally installed. You must provide the path to a prepared STAR genome directory. pypy and pigz are preferred for performance but not required. Note that STAR uses a lot of memory, e.g. about 25 GB for the human genome, so be sure to run this on a suitable computer.

Options

• -n [length]: extract leading UMI of specified length (default 5)
• -g [length]: discard specified number of bases after UMI (default 3)
• -t [length]: truncate reads to specified length
• -d: skip duplicate marking

slurm_align_smart-3seq.sh

Version of align_smart-3seq.sh to use on computing clusters powered by SLURM (submits a job array).

quasi-align_smart-3seq.sh and slurm_quasi-align_smart-3seq.sh

Experimental pipeline using Salmon to quasi-align the data with special options that make sense for 3SEQ. This doesn't seem to get good results, though.

Read-counting pipeline

make_expression_table.R

Makes a table of gene expression, including read counts, TPM, and optionally DESeq2 rlog, from a list of BAM files. rlog may take a long time with large numbers of samples so this can be disabled with --no-rlog.

Quality-control scripts

categorize_alignments.py

Categorizes alignments by their positions relative to annotated genes.

qc_3seq.R

Makes graphs and tables from the output (log files) of align_smart-3seq.sh or equivalent.

graph_trimmed_lengths.R

Makes graphs of trimmed insert lengths from the log files of umi_homopolymer.py.

aggregate_star_logs.py

Collects the log files from STAR alignments into a simple spreadsheet.

trim_fastqc.sh

Wrapper to run FastQC after trimming the data with umi_homopolymer.py.

Miscellaneous

find_reference_homopolymers.py

Uses a very (too?) simple algorithm to find sites in the reference sequence that might attract off-target priming with the oligo(dT) anchor.

blacklist_alignments.py

Marks or removes reads from a BAM file that are inside unwanted regions in a BED file. Not sure if this is useful.