Pheniqs

Pheniqs performs probabilistic decoding of combinatorial sequencing barcodes to associate next-generation sequencing reads with their originating samples, cells, or molecules.

Key Features:

• Probabilistic Decoding: Employs Bayesian methods to compute full posterior decoding error probabilities for observed barcodes using basecalling quality scores and prior distributions as an alternative to minimum edit distance or simple maximum likelihood estimation.
• Flexible Configuration: Supports arbitrarily complex barcoding designs and configurable arrangements of combinatorial barcodes.
• Scalability and Performance: Scales linearly with core count and can classify over 11 billion reads in approximately 1 hour and 15 minutes while using less than 50 megabytes of memory.
• Confidence Scoring: Calculates overall confidence scores by taking the product of posterior probabilities for multiple independent barcodes, enabling tuning of decoding sensitivity via confidence thresholds.
• Integration and Extensibility: Supports POSIX streams and multiple sequencing formats and is extensible with alternative decoders and new error models.
• Reproducibility and Benchmarking: Preserves decoding quality scores and provides tunable control over classification confidence to support reproducibility and community benchmarking.

Scientific Applications:

• Systems Biology: Decodes combinatorial index tags to link sequencing reads to samples, cells, or molecular origins in systems biology experiments.
• Combinatorial Indexing: Enables accurate decoding of complex combinatorial barcode schemes used to tag biological material in multiplexed assays.
• High-throughput Genomics: Provides scalable barcode classification for deep sequencing studies and genomic facilities requiring processing of billions of reads.

Methodology:

Computes Bayesian posterior decoding probabilities integrating basecalling quality scores and prior distributions, multiplies posteriors for independent barcodes to obtain overall confidence, and supports alternative decoders and error models.