acdc

acdc detects contamination in single-cell sequencing (SCS) data by combining reference-based and reference-free computational methods to identify known and novel foreign DNA and support high-quality genomic assemblies.

Key Features:

• Reference-free contaminant identification: Employs a reference-free strategy to detect contaminants from organisms lacking closely related reference species, avoiding sole reliance on database alignment or marker gene searches.
• Dimensionality reduction and clustering: Performs fast, non-linear dimensionality reduction on oligonucleotide signatures followed by clustering algorithms that automatically estimate the number of clusters to distinguish taxa and contaminants.
• Hybrid supervised/unsupervised methodology: Combines 16S rRNA gene prediction and ultrafast exact alignment for sequence classification with subsequent reference-free machine learning inspection.
• Statistical confidence assessment: Applies bootstrapping to provide confidence values for clustering results and contaminant calls.

Scientific Applications:

• Single-cell genome contamination screening: Detects and differentiates contaminant DNA in single-cell sequencing (SCS) datasets to reduce erroneous biological interpretations.
• Genomic assembly quality assurance: Improves quality control of genomic sequence data and assemblies prior to submission to public databases.
• Detection of novel contaminants: Identifies contaminants from organisms lacking close references, enabling discovery of novel or uncharacterized taxa.

Methodology:

Initial sequence classification using 16S rRNA gene prediction and ultrafast exact alignment; non-linear dimensionality reduction on oligonucleotide signatures; clustering with automatic cluster-number estimation; bootstrapping to assess confidence.