RemoveSmartBell

Basic Usage

removesmartbell.sh in=<input> out=<output> split=t

Input may be fasta or fastq, compressed or uncompressed (not H5 files).

Parameters

Parameters control adapter detection, processing modes, and output formatting.

Core Parameters

in=file: Specify the input file, or stdin. Can use # notation for paired files (e.g., reads#.fq becomes reads1.fq and reads2.fq).
in2=file: Specify the second input file for paired data.
out=file: Specify the output file, or stdout. Can use # notation for paired output files.
adapter=string: Specify the adapter sequence. Default is normal SmrtBell adapter sequence (ATCTCTCTCTTTTCCTCCTCCTCCGTTGTTGTTGTTGAGAGAGAT).
split=t: Processing mode: t=splits reads at adapters into separate contigs, f=masks adapters with X symbols but keeps reads intact. Default: true

Alignment Parameters

minratio=0.31: Minimum alignment score ratio for adapter detection. At 250bp reads, approximately 0.01% false-positive and 94% true-positive rate.
suspectratio=0.85: Ratio threshold for suspect alignments that may be confirmed by nearby adapters or secondary alignment methods.
usealtmsa=t: Enable alternate multi-state alignment algorithm for improved sensitivity. Uses MultiStateAligner9PacBioAdapter2 as backup method.
plusonly=f: Only search for adapters in forward orientation. When true, disables reverse complement search.
minusonly=f: Only search for adapters in reverse complement orientation. When true, disables forward search.

Quality Control Parameters

mincontig=50: Minimum contig length to retain after splitting. Shorter sequences are discarded.
reads=unlimited: Maximum number of reads to process. Supports K/M/G suffixes (e.g., reads=1M).
maxreads=unlimited: Alias for reads parameter.

Processing Parameters

threads=auto: Number of processing threads. Use 'auto' to detect available processors automatically.
overwrite=t: Allow overwriting of existing output files.
append=f: Append to existing output files instead of overwriting.
verbose=f: Print additional processing information and statistics.

Common Parser Parameters

path=: Set the path for temporary files and working directory.
tempdir=: Specify temporary directory for intermediate files.
root=: Root directory for relative file paths.

Examples

Basic Adapter Removal

removesmartbell.sh in=pacbio_reads.fq out=clean_reads.fq split=t

Removes Smart Bell adapters from PacBio reads, splitting reads at adapter locations.

Mask Adapters Instead of Splitting

removesmartbell.sh in=pacbio_reads.fq out=masked_reads.fq split=f

Masks adapter sequences with X symbols but keeps reads intact as single sequences.

Custom Adapter Sequence

removesmartbell.sh in=reads.fq out=clean.fq adapter=CUSTOMADAPTERSEQUENCE

Uses a custom adapter sequence instead of the default Smart Bell adapter.

High Sensitivity Processing

removesmartbell.sh in=reads.fq out=clean.fq minratio=0.25 usealtmsa=t

Increases sensitivity by lowering alignment threshold and enabling alternate alignment algorithm.

Paired-End Processing

removesmartbell.sh in=reads#.fq out=clean#.fq

Processes paired-end files reads1.fq and reads2.fq, outputting to clean1.fq and clean2.fq.

Algorithm Details

Alignment Strategy

RemoveSmartBell uses MultiStateAligner algorithms specifically designed for PacBio adapter detection:

Primary Algorithm: MultiStateAligner9PacBioAdapter for local alignment scoring and adapter detection
Backup Algorithm: MultiStateAligner9PacBioAdapter2 (when usealtmsa=true) for secondary scoring when primary alignments are ambiguous
Dual-Direction Search: Searches for adapters in both forward and reverse complement orientations unless restricted by plusonly/minusonly parameters

Scoring and Thresholds

The algorithm implements a multi-threshold scoring system based on Smith-Waterman alignment scores:

Primary Threshold (minratio=0.31): minSwScore = maxSwScore * 0.31, where maxSwScore = msa.maxQuality(query_length), achieves ~0.01% false positive rate at 250bp read length
Suspect Threshold (suspectratio=0.85): minSwScoreSuspect calculated as maxSwScore * min(0.31 * 0.85, 0.31 - ((1-0.85) * 0.2)) for intermediate confidence detections
Locality-Based Confirmation: Suspect alignments within suspectDistance=100bp of confirmed adapters are accepted using lastConfirmed and lastSuspect position tracking
Look-Ahead Validation: When array.length-stop > window, performs additional msa.fillAndScoreLimited() call on downstream window-sized region

Processing Strategy

The algorithm processes reads using a sliding window approach:

Window Size: window = query1.length * 2.5 + 10 (calculated as int)(query1.length*2.5f+10)
Stride: stride = query1.length * 0.95 (calculated as int)(query1.length*0.95f) to ensure overlap detection
Padding: npad=35bp N-padding on sequence ends using npad() method to handle boundary conditions
Threading: ProcessThread[] array with THREADS=Shared.LOGICAL_PROCESSORS for parallel processing

Output Modes

Two primary processing modes are available:

Split Mode (split=t): Divides reads at adapter locations, creating multiple shorter contigs from each original read. Discards resulting sequences shorter than mincontig threshold
Mask Mode (split=f): Replaces adapter sequences with X symbols while maintaining read integrity and length

Performance Characteristics

Implementation characteristics based on source code analysis:

Memory Usage: Base allocation z="-Xmx400m" with Shared.capBufferLen(20) buffer length capping
Throughput: ConcurrentReadInputStream and ConcurrentReadOutputStream with configurable buffer sizes
Accuracy: Tuned with MINIMUM_ALIGNMENT_SCORE_RATIO=0.31f and SUSPECT_RATIO=0.85F constants
Scalability: ProcessThread.run() loop processes ListNum<Read> batches for concurrent execution

Statistics and Output

RemoveSmartBell provides comprehensive statistics on adapter detection performance:

Processing Statistics

Reads Processed: Total input reads and base count
Good vs Bad Reads: Counts of reads with and without adapters
Adapter Counts: Separate counts for forward and reverse complement adapters
Adapter Density: Adapters per megabase for contamination assessment
Output Summary: Final read and base counts after processing

Accuracy Metrics

For synthetic data with known adapter positions:

True Positive: Correctly identified adapters
False Positive: Incorrectly identified adapters
True Negative: Correctly identified clean regions
False Negative: Missed adapters

Support

For questions and support:

Email: bbushnell@lbl.gov
Documentation: bbmap.org