RNA-Bloom is a fast and memory-efficient de novo transcript sequence assembler. It is designed for the following sequencing data types:

• single-end/paired-end bulk RNA-seq (strand-specific/agnostic)
• paired-end single-cell RNA-seq (strand-specific/agnostic)
• long-read RNA-seq (ONT cDNA/direct RNA, PacBio cDNA)

Written by Ka Ming Nip

2018-present Canada’s Michael Smith Genome Sciences Centre, BC Cancer

---

Dependency 📌

• Java SE Development Kit (JDK) 11 (JDK 17 is slightly faster)

• External software used:

software	short reads	long reads
minimap2 >=2.22	required	required
Racon	not used	required
ntCard >=1.2.1	required	required

Their executables must be accessible from your PATH!

Installation 🔧

RNA-Bloom can be installed in two ways:

(A) install with conda or mamba:

conda install -c bioconda rnabloom

mamba install -c bioconda rnabloom

All dependent software (listed above) will be installed. RNA-Bloom can be run as rnabloom ...

(B) download from GitHub:

1. Download the binary tarball rnabloom_vX.X.X.tar.gz from the releases section.
2. Extract the downloaded tarball with the command:

   tar -zxf rnabloom_vX.X.X.tar.gz

   RNA-Bloom can be run as java -jar /path/to/RNA-Bloom.jar ...

Quick Start for Short Reads

Input reads must be in either FASTQ or FASTA format and may be compressed with GZIP.

Note that -left, -right, -sef, and -ser can accept multiple file paths separated by the whitespace character.

(A) assemble bulk RNA-seq data:

• paired-end reads only

  • when left reads are sense and right reads are antisense, use -revcomp-right to reverse-complement right reads
  • when left reads are antisense and right reads are sense, use -revcomp-left to reverse-complement left reads
  • for non-stranded data, use either -revcomp-right or -revcomp-left

    java -jar RNA-Bloom.jar -left LEFT.fastq -right RIGHT.fastq -revcomp-right -t THREADS -outdir OUTDIR

• single-end reads only

  • use -sef for forward reads and -ser for reverse reads

    java -jar RNA-Bloom.jar -sef SE.fastq -t THREADS -outdir OUTDIR

• paired-end and single-end reads

  java -jar RNA-Bloom.jar -left LEFT.fastq -right RIGHT.fastq -revcomp-right -sef SE.fastq -t THREADS -outdir OUTDIR

final output files:

file name	description
rnabloom.transcripts.fa	assembled transcripts longer than length threshold (default: 200)
rnabloom.transcripts.short.fa	assembled transcripts shorter than length threshold
rnabloom.transcripts.nr.fa	assembled transcripts with redundancy reduced

(B) assemble multi-sample RNA-seq data with pooled assembly mode:

java -jar RNA-Bloom.jar -pool READSLIST.txt -revcomp-right -t THREADS -outdir OUTDIR

This is especially useful for single-cell datasets. RNA-Bloom was tested on Smart-seq2 and SMARTer datasets. It is not supported for long-read data (-long) at this time.

file format for the -pool option:

This is a tabular file that describes the read file paths for all cells/samples to be used pooled assembly.

• Column header is on the first line, leading with #
• Columns are separated by space/tab characters
• Each sample can have more than one lines; lines sharing the same name will be grouped together during assembly

column	description
name	sample name
left	path to one left read file
right	path to one right read file
sef	path to one single-end forward read file
ser	path to one single-end reverse read file

(i) paired-end reads only:

Only name, left, and right columns are specified for a total of 3 columns. The legacy header-less tri-column format is still supported.

#name left right
cell1 /path/to/cell1/left.fastq /path/to/cell1/right.fastq
cell2 /path/to/cell2/left.fastq /path/to/cell2/right.fastq
cell3 /path/to/cell3/left.fastq /path/to/cell3/right.fastq

(ii) paired and unpaired reads:

In addition to name, left, and right columns, either sef, ser or both are specified for a total of 4~5 columns.

#name left right sef ser
cell1 /path/to/cell1/left.fastq /path/to/cell1/right.fastq /path/to/cell1/sef.fastq /path/to/cell1/ser.fastq
cell2 /path/to/cell2/left.fastq /path/to/cell2/right.fastq /path/to/cell2/sef.fastq /path/to/cell2/ser.fastq
cell3 /path/to/cell3/left.fastq /path/to/cell3/right.fastq /path/to/cell3/sef.fastq /path/to/cell3/ser.fastq

final output files per cell:

file name	description
rnabloom.transcripts.fa	assembled transcripts longer than length threshold (default: 200)
rnabloom.transcripts.short.fa	assembled transcripts shorter than length threshold
rnabloom.transcripts.nr.fa	assembled transcripts with redundancy reduced

(C) strand-specific assembly:

java -jar RNA-Bloom.jar -stranded ...

The -stranded option indicates that input reads are strand-specific.

Strand-specific reads are typically in the F2R1 orientation, where /2 denotes left reads in forward orientation and /1 denotes right reads in reverse orientation.

Configure the read file paths accordingly for bulk RNA-seq data and indicate read orientation:

-stranded -left /path/to/reads_2.fastq -right /path/to/reads_1.fastq -revcomp-right

and for scRNA-seq data:

cell1 /path/to/cell1/reads_2.fastq /path/to/cell1/reads_1.fastq

(D) reference-guided assembly:

java -jar RNA-Bloom.jar -ref TRANSCRIPTS.fasta ...

The -ref option specifies the reference transcriptome FASTA file for guiding short-read assembly. It is not supported for long-read data (-long) at this time.

Quick Start for Long Reads

It is strongly recommended to trim adapters in your reads before assembly. For example, see Porechop for more information.

Input reads must not have purely integer IDs (e.g. 1, 2, 3), which could be in conflict with RNA-Bloom’s sequence IDs. Please rename your read IDs (with seqtk rename) if necessary.

Note that -long, -sef, and -ser can accept multiple file paths separated by the whitespace character.

(A) assemble long-read cDNA sequencing data:

Default presets for -long are intended for ONT data. Please add the -lrpb flag for PacBio data.

java -jar RNA-Bloom.jar -long LONG.fastq -t THREADS -outdir OUTDIR

Input reads are expected to be in a mix of both forward and reverse orientations.

Options -pool and -ref are not supported for long-read data at this time.

(B) assemble nanopore direct RNA sequencing data:

java -jar RNA-Bloom.jar -long LONG.fastq -stranded -t THREADS -outdir OUTDIR

Input reads are expected to be only in the forward orientation.

By default, uracil (U) is written as T. Use the -uracil option to write U instead of T in the output assembly.

ntCard v1.2.1 supports uracil in reads.

(C) assemble long-read sequencing data with short-read polishing:

cDNA data:

java -jar RNA-Bloom.jar -long LONG.fastq -sef SHORT.fastq -t THREADS -outdir OUTDIR

direct RNA data:

java -jar RNA-Bloom.jar -stranded -long LONG.fastq -sef SHORT_FORWARD.fastq -ser SHORT_REVERSE.fastq -t THREADS -outdir OUTDIR

final output files:

file name	description
rnabloom.transcripts.fa	assembled transcripts longer than min. length threshold (default: 200)
rnabloom.transcripts.short.fa	assembled transcripts shorter than min. length threshold

General Settings :gear:

(A) set Bloom filter sizes automatically:

If ntcard is found in your PATH, then the -ntcard option is automatically turned on to count the number of unique k-mers in your reads.

java -jar RNA-Bloom.jar -fpr 0.01 ...

This sets the size of Bloom filters automatically to accommodate a false positive rate (FPR) of ~1%.

Alternatively, you can specify the exact number of unique k-mers:

java -jar RNA-Bloom.jar -fpr 0.01 -nk 28077715 ...

This sets the size of Bloom filters automatically to accommodate 28,077,715 unique k-mers for a FPR of ~1%.

As a rule of thumb, a lower FPR may result in a better assembly but requires more memory for a larger Bloom filter.

(B) set the total size of Bloom filters:

java -jar RNA-Bloom.jar -mem 10 ...

This sets the total size to 10 GB. If neither -nk, -ntcard, or -mem are used, then the total size is configured based on the size of input read files.

(C) stop at an intermediate stage:

java -jar RNA-Bloom.jar -stage N ...

N	short reads	long reads
1	construct graph	construct graph
2	assemble fragments	correct reads
3	assemble transcripts	assemble transcripts

This is a very useful option if you only want to assemble fragments or correct long reads (ie. with -stage 2)!

(D) list all available options in RNA-Bloom:

java -jar RNA-Bloom.jar -help

(E) limit the size of Java heap:

java -Xmx2g -jar RNA-Bloom.jar ...

or if you installed with conda:

export JAVA_TOOL_OPTIONS="-Xmx2g"
rnabloom ...

This limits the maximum Java heap to 2 GB with the -Xmx option. Note that java options has no effect on Bloom filter sizes.

See documentation for other JVM options.

Implementation 📝

RNA-Bloom is written in Java with Apache NetBeans IDE. It uses the following libraries:

• Apache Commons CLI
• JGraphT
• Smile

Citing RNA-Bloom

If you use RNA-Bloom in your work, please cite our manuscript(s).

Long-read RNA-seq assembly:

Ka Ming Nip, Saber Hafezqorani, Kristina K. Gagalova, Readman Chiu, Chen Yang, René L. Warren, and Inanc Birol. Reference-free assembly of long-read transcriptome sequencing data with RNA-Bloom2. Nature Communications. 2023 May 22;14(1):2940. doi: 10.1038/s41467-023-38553-y

Short-read RNA-seq assembly:

Ka Ming Nip, Readman Chiu, Chen Yang, Justin Chu, Hamid Mohamadi, René L. Warren, and Inanc Birol. RNA-Bloom enables reference-free and reference-guided sequence assembly for single-cell transcriptomes. Genome Research. 2020 Aug;30(8):1191-1200. doi: 10.1101/gr.260174.119. Epub 2020 Aug 17.

---