USEARCH

USEARCH performs sensitive, high-throughput sequence searching and error reduction for NGS and amplicon data to support homology detection and functional inference from large sequence databases.

Key Features:

• High-throughput sensitive search: Performs fast, non-profile-based homology searches against large sequence databases for functional inference.
• Error-correction strategies: Implements three independent error-reduction approaches: filtering reads by expected number of errors, assembling overlapping read pairs, and leveraging unique sequence abundances in amplicon reads.
• Low-coverage discrimination: Distinguishes true biological variation from sequencing errors, including under low-coverage conditions.
• Benchmarking against other methods: Has been systematically compared to CS-BLAST, HHSEARCH, PHMMER, NCBI-BLAST, UBLAST, and FASTA using challenging datasets.
• Domain-architecture evaluations: Evaluations used protein domain architecture schemes such as PFAM+Clan, SCOP/Superfamily, and CATH/Gene3D.
• Speed–accuracy trade-off: Offers significant speed advantages relative to profile-based methods while exhibiting lower remote-homolog detection accuracy than top profile-based tools.

Scientific Applications:

• Homology detection and functional annotation: Infers function from sequences deposited in public databases via non-profile homology searches.
• NGS and amplicon error reduction: Improves accuracy of amplicon and NGS datasets by applying expected-error filtering, read-pair assembly, and abundance-based correction.
• Low-coverage variant discrimination: Facilitates separation of sequencing errors from true biological variation in low-coverage data.
• Method benchmarking: Serves as a comparator in systematic evaluations of profile-based and non-profile-based homology search tools.

Methodology:

Uses three explicit error-correction methods—filtering reads by expected error counts, assembling overlapping read pairs, and leveraging unique sequence abundances in amplicon reads—and performs non-profile-based homology searches benchmarked against CS-BLAST, HHSEARCH, PHMMER, NCBI-BLAST, UBLAST, and FASTA using datasets based on PFAM+Clan, SCOP/Superfamily, and CATH/Gene3D.