BLAST NCBI Search: A Practical Guide to Sequence Similarity
By Dr. Zubair Khalid, DVM, MS, PhD ·
Introduction to BLAST and NCBI Search
What is BLAST?
The Basic Local Alignment Search Tool (BLAST) is a family of algorithms designed to compare a query sequence against a database of sequences and identify regions of local similarity. Developed by Stephen Altschul and colleagues at the National Center for Biotechnology Information (NCBI) in 1990, BLAST implements a heuristic approach that sacrifices exhaustive optimality for computational speed, making it feasible to search databases containing billions of residues in seconds.
The fundamental principle underlying BLAST is local alignment. Unlike global alignment methods such as Needleman-Wunsch, which align entire sequences end-to-end, local alignment identifies short, high-scoring regions of similarity that may be embedded within otherwise divergent sequences. This is biologically appropriate because protein domains, regulatory motifs, and functional modules are often conserved independently of surrounding sequence context.
BLAST works by first identifying short "words" (typically 3 residues for proteins, 11 nucleotides for DNA) that score above a threshold when matched between query and database sequences. These word hits are then extended in both directions to generate ungapped alignments, and finally, gapped extensions are attempted for promising candidates. The statistical significance of each alignment is computed using an extreme value distribution, which models the distribution of scores expected from random matches in a database of a given size.
The statistical framework of BLAST is grounded in the work of Samuel Karlin and Stephen Altschul, who demonstrated that the scores of optimal local alignments between random sequences follow a Gumbel distribution. This allows calculation of an Expectation value (E-value), which represents the number of alignments with a given score (or better) that would be expected to occur by chance in a database search. An E-value of 0.001 means that one such alignment would be expected by chance in 1,000 database searches of the same size.
The NCBI BLAST Portal
The NCBI BLAST portal (https://blast.ncbi.nlm.nih.gov) serves as the primary web interface for BLAST searches. It provides access to all major BLAST programs, a comprehensive set of sequence databases, and a suite of visualization and analysis tools. The portal is designed to accommodate users ranging from occasional researchers to bioinformaticians performing high-throughput analyses.
The interface organizes searches by query type and program. The main page offers quick links to Nucleotide BLAST (blastn), Protein BLAST (blastp), and Translated BLAST (blastx, tblastn, tblastx). For each search, the user specifies a query sequence (either by pasting FASTA-formatted text or providing an accession number), selects a database, and adjusts parameters. The portal also provides specialized search interfaces for particular use cases, such as primer alignment, domain searching, and whole-genome comparisons.
A critical feature of the NCBI portal is the ability to restrict searches by organism. This is implemented through the "Organism" field, which accepts taxonomic names or NCBI Taxonomy IDs. The portal automatically applies the appropriate taxonomic restriction to the selected database, which can dramatically reduce search time and improve signal-to-noise ratio when working with a known species.
BLAST Algorithm Variants and Their Uses
Nucleotide BLAST (blastn)
The blastn program compares a nucleotide query against a nucleotide database. It is optimized for finding high-identity matches between DNA sequences, such as identifying an unknown sequence, confirming the presence of a gene in a genome, or locating a sequence within a larger contig.
blastn uses a word size of 11 by default, which provides high sensitivity for sequences that share at least 90% identity over short regions. For cross-species comparisons where sequences may be more divergent, the word size can be reduced to 7, which increases sensitivity at the cost of computational speed. The scoring scheme for blastn is typically a simple match/mismatch system, with the default being +2 for a match and -3 for a mismatch, though these values can be adjusted.
A specialized variant of blastn is the "megablast" algorithm, which uses a word size of 28 and is optimized for comparing highly similar sequences, such as those from the same species or closely related strains. Megablast is substantially faster than standard blastn and is the default choice for many genome assembly and annotation tasks. Conversely, "discontiguous megablast" uses a word size of 11 with non-contiguous word matching, making it suitable for cross-species comparisons where sequence identity is lower.
Protein BLAST (blastp)
The blastp program compares a protein query against a protein database. It is the most commonly used BLAST variant for functional annotation, because protein sequences are more conserved than the underlying nucleotide sequences due to the degeneracy of the genetic code and the selective pressure to maintain protein structure and function.
blastp uses a word size of 3 by default and employs a substitution matrix, most commonly BLOSUM62, to score amino acid matches and mismatches. The BLOSUM62 matrix was derived from alignments of protein blocks with at least 62% identity and provides a good balance between sensitivity for distant homologs and specificity for close relatives. For more divergent searches, the BLOSUM45 matrix (derived from alignments with at least 45% identity) can be used, while BLOSUM80 is appropriate for very closely related sequences.
The blastp algorithm also includes a composition-based statistics adjustment that corrects for the amino acid composition of the query. This adjustment is important because proteins with biased composition (e.g., proline-rich or glycine-rich proteins) can produce artificially high scores against unrelated sequences. The default "composition-adjusted" scoring is recommended for most searches.
Translated BLAST (blastx, tblastn, tblastx)
Translated BLAST programs compare sequences in different alphabets by translating nucleotide sequences into protein sequences in all six reading frames. These programs are essential for identifying protein-coding regions in unannotated nucleotide sequences and for detecting homology that is invisible at the nucleotide level due to synonymous codon changes.
blastx translates a nucleotide query in all six reading frames and compares the resulting protein sequences against a protein database. This is particularly useful for analyzing newly sequenced genomic DNA or ESTs (expressed sequence tags) where the coding region is unknown. For example, a researcher who has sequenced a cDNA clone of unknown function can use blastx to identify the encoded protein by searching against the non-redundant protein database. The six-frame translation ensures that the correct reading frame is captured regardless of where the coding sequence begins.
tblastn is the inverse of blastx: it translates a nucleotide database in all six reading frames and compares a protein query against these translations. This is valuable for finding protein-coding regions in genomic DNA, especially in organisms with intron-rich genomes where the coding sequence is fragmented. A researcher studying a conserved protein in a newly sequenced genome can use tblastn to identify the exons encoding the protein, even if the gene annotation is incomplete.
tblastx translates both the query and the database in all six reading frames and compares the resulting protein sequences. This is computationally expensive and is rarely used for large database searches. It is most appropriate for comparing two small sets of sequences where nucleotide-level similarity is too low to be detected by blastn but protein-level conservation is expected. For most applications, blastx or tblastn provides sufficient sensitivity at a fraction of the computational cost.
Selecting the Appropriate Database
Non-redundant (nr) Database
The non-redundant (nr) database is the default choice for most BLAST searches and is the largest protein database at NCBI. Despite its name, nr is not strictly non-redundant; it is a composite of sequences from GenBank, RefSeq, Swiss-Prot, PDB, and other sources, with identical sequences collapsed into single entries. The database contains over 500 million protein sequences as of 2024, representing all major taxonomic groups.
The primary advantage of nr is its comprehensiveness. A search against nr will identify homologs from any organism, including those with minimal sequence data. This makes it the database of choice for functional annotation of unknown sequences and for identifying the taxonomic distribution of a protein family. The primary disadvantage is that the database is heavily biased toward model organisms and medically important species, which can skew the statistical significance of hits from underrepresented taxa.
For nucleotide searches, the corresponding database is "nt" (nucleotide collection), which contains all GenBank nucleotide sequences. The nt database is similarly comprehensive but is dominated by sequences from a relatively small number of well-studied species.
Reference Sequence (RefSeq) Database
The RefSeq database provides a curated, non-redundant set of reference sequences for major organisms. Each RefSeq entry is annotated with functional information, and the database is maintained by NCBI staff with input from the research community. RefSeq entries are identified by their accession prefix: NC_ for complete chromosomes, NM_ for mRNA, NP_ for protein, and NG_ for genomic regions that cannot be fully assembled.
The advantage of RefSeq over nr is the quality of annotation and the reduced redundancy. Searching against RefSeq proteins (the "refseq_protein" database) will return well-annotated entries with functional descriptions, which is valuable for interpreting the biological significance of hits. The disadvantage is that RefSeq may lack sequences from poorly studied organisms or recently discovered genes. For a comprehensive search, it is often advisable to search nr first and then confirm hits against RefSeq.
Genome and Transcriptome Databases
NCBI provides organism-specific genome databases that contain all sequences from a particular species or taxonomic group. These are accessed through the "Genome" section of the BLAST portal or by specifying the organism in the search parameters. Genome databases are essential for mapping a query sequence to a specific genomic location, identifying splice variants, or comparing gene content between strains.
For transcriptome analysis, the "EST" (Expressed Sequence Tag) and "TSA" (Transcriptome Shotgun Assembly) databases contain short cDNA sequences and assembled transcripts, respectively. These databases are useful for confirming gene expression, identifying alternative splice forms, and characterizing genes in organisms without complete genome assemblies. The TSA database is particularly valuable for non-model organisms, as it contains transcriptome assemblies from a wide range of species.
The choice of database should be guided by the biological question. For identifying the function of an unknown protein, nr is appropriate. For mapping a gene to a specific genome, the organism-specific genome database is necessary. For confirming expression or identifying splice variants, EST or TSA databases are suitable.
Setting BLAST Parameters for Optimal Results
E-value and Bit Score
The E-value is the most frequently misinterpreted parameter in BLAST searches. An E-value is not a probability that a hit is a false positive; rather, it is the expected number of alignments with a score at least as high as the observed score that would occur by chance in a database of the same size. An E-value of 1 means that one such alignment would be expected by chance in a database of that size. Lower E-values indicate more significant matches.
The bit score is a normalized version of the raw alignment score that is independent of the scoring system and database size. Bit scores are calculated as:
S' = (λS - ln K) / ln 2
where S is the raw score, λ and K are statistical parameters that depend on the scoring system and sequence composition, and S' is the bit score. Bit scores allow comparison of alignment significance across different scoring matrices and database sizes.
The default E-value threshold for most BLAST searches is 10, which is quite permissive. For most applications, a threshold of 0.001 or lower is appropriate. However, the optimal threshold depends on the purpose of the search. For identifying close homologs in a well-annotated genome, a stringent threshold of 1e-10 is reasonable. For detecting distant homologs in a large database, a more permissive threshold may be necessary, but the results should be interpreted with caution.
Word Size and Gap Penalties
The word size determines the minimum length of the initial seed match that triggers an alignment extension. Larger word sizes increase speed but decrease sensitivity for divergent sequences. For blastn, the default word size is 11, which is appropriate for sequences with at least 90% identity. Reducing the word size to 7 increases sensitivity for cross-species comparisons but increases search time by approximately an order of magnitude.
For blastp, the default word size is 3, which provides a balance between sensitivity and speed. Reducing the word size to 2 increases sensitivity for very divergent proteins but is computationally expensive. The word size for blastp is effectively constrained by the amino acid alphabet: a word size of 2 represents 400 possible dipeptides, while a word size of 3 represents 8,000 possible tripeptides.
Gap penalties control the cost of introducing insertions and deletions into an alignment. The default gap penalties for blastp are an 11-residue gap opening penalty and a 1-residue gap extension penalty. These values were optimized on benchmark datasets and work well for most protein families. For sequences with unusual gap patterns, such as proteins with variable-length linker regions, adjusting the gap extension penalty may improve alignment quality.
Low Complexity Filtering
Low complexity regions (LCRs) are sequences with biased amino acid or nucleotide composition, such as poly-proline tracts, coiled-coil regions, or simple nucleotide repeats. These regions produce high-scoring alignments with unrelated sequences that share similar composition, generating spurious hits that obscure true homologs.
The default behavior of BLAST is to filter low complexity regions using the SEG algorithm for proteins and the DUST algorithm for nucleotides. Filtering replaces low complexity regions with "N" or "X" characters, which do not participate in alignment. This is generally recommended for standard searches, as it dramatically reduces false positives.
However, filtering can mask biologically meaningful features. For example, a protein with a functionally important proline-rich domain will have that domain masked, potentially missing true homologs that share the domain. In such cases, the filter can be disabled, but the results should be interpreted with caution. A useful compromise is to perform the search with filtering enabled and then repeat the search with filtering disabled for the top hits to determine whether the low complexity region is contributing to the alignment.
Interpreting BLAST Output and Alignment Statistics
Understanding the Hit Table
The BLAST results page presents a hit table listing database sequences that produced significant alignments. Each row displays the accession number, description, bit score, E-value, and the percentage of query coverage and identity. The hit table is sorted by E-value, with the most significant hits at the top.
The accession number identifies the database sequence and provides a link to the full record. The description typically includes the gene name, organism, and functional annotation. The bit score and E-value provide measures of statistical significance, with lower E-values and higher bit scores indicating more significant matches.
The query coverage percentage indicates the fraction of the query sequence that is included in the alignment. A hit with high identity but low coverage may represent a shared domain rather than a full-length homolog. Conversely, a hit with low identity but high coverage may represent a distantly related homolog with conserved overall structure. Both metrics must be considered together to interpret the biological significance of a hit.
Reading the Alignment
The alignment view displays the pairwise alignment between the query and each database sequence. The alignment shows the query sequence on top and the database sequence on the bottom, with a middle line indicating identical residues (shown as the amino acid or nucleotide letter) and conservative substitutions (shown as "+" for proteins).
For protein alignments, the "+" symbol indicates a positive-scoring substitution according to the BLOSUM62 matrix. For example, a leucine-to-isoleucine substitution would be scored as positive because these residues are similar in size and hydrophobicity. The alignment also shows gaps, represented by dashes, which indicate insertions or deletions relative to the other sequence.
The alignment frame includes the start and end positions of the aligned region in both sequences, as well as the frame for translated searches. For blastx and tblastn, the reading frame is indicated (e.g., "+1" or "-3"), which is essential for identifying the correct open reading frame in the nucleotide sequence.
Graphical Summary
The graphical summary provides a visual representation of the alignment coverage across the query sequence. Each hit is shown as a colored bar positioned according to its alignment coordinates on the query. The color of the bar indicates the bit score, with red representing the highest scores and black representing the lowest.
The graphical summary is particularly useful for identifying multi-domain proteins. A query that produces multiple distinct bars at different positions may contain several conserved domains, each matching different database sequences. This pattern is common for modular proteins such as transcription factors containing both DNA-binding and activation domains.
The graphical summary also helps identify partial hits. A bar that covers only a portion of the query sequence indicates that the database sequence shares similarity with only that region. This could represent a shared domain, a truncated protein, or a pseudogene.
Advanced BLAST Search Strategies
PSI-BLAST for Distant Homologs
Position-Specific Iterated BLAST (PSI-BLAST) is an extension of blastp that detects distant homologs by constructing a position-specific scoring matrix (PSSM) from the results of an initial search. The PSSM captures the amino acid preferences at each position in the query, allowing detection of sequences that share conserved features despite low overall identity.
The PSI-BLAST algorithm proceeds in iterations. The first iteration is a standard blastp search against the specified database. Sequences that pass a threshold (default E-value of 0.005) are used to construct a PSSM. In the second iteration, the PSSM is used to search the database again, which typically identifies additional homologs that were not detected in the first iteration. These new hits are added to the PSSM, and the process is repeated until convergence or a specified number of iterations.
PSI-BLAST is particularly powerful for identifying members of large protein families with diverse sequences. For example, a search with a single kinase domain sequence using PSI-BLAST can identify hundreds of kinases across the tree of life, whereas a standard blastp search might identify only the closest relatives. The Conserved Sequence analysis enabled by PSI-BLAST is essential for understanding the evolutionary relationships within such families.
However, PSI-BLAST is susceptible to contamination by spurious hits. Once a false positive enters the PSSM, it can propagate through subsequent iterations, recruiting additional false positives. This is particularly problematic for sequences with low complexity regions or for searches against databases with many paralogs. The Paralog and Ortholog distinction is critical here: PSI-BLAST will happily include paralogs in the PSSM, which is appropriate for identifying family members but can obscure the orthology relationships that are often the biological question of interest.
PHI-BLAST for Motif Patterns
Pattern Hit Initiated BLAST (PHI-BLAST) combines a user-specified protein motif with a BLAST search. The motif is defined using a regular expression pattern, such as [LIVM]-x(2)-G-x(3)-[ST]-x(2)-[LIVM], which specifies the allowed residues at each position. PHI-BLAST first searches the database for sequences containing the motif and then performs a blastp search restricted to those sequences.
PHI-BLAST is valuable for identifying proteins that share a specific functional motif but are otherwise divergent. For example, a researcher studying a family of DNA-binding proteins could define the DNA-binding motif and use PHI-BLAST to identify all proteins containing that motif, regardless of their overall sequence similarity.
The motif must be carefully defined to balance specificity and sensitivity. A motif that is too restrictive will miss true members of the family, while a motif that is too permissive will return many false positives. The pattern syntax follows the PROSITE convention, with brackets indicating allowed residues, braces indicating disallowed residues, and "x" indicating any residue.
Common Pitfalls and Troubleshooting in BLAST Searches
Misleading E-values
The most common error in BLAST interpretation is treating E-values as probabilities. An E-value of 0.05 does not mean there is a 5% chance the hit is a false positive. Rather, it means that 0.05 such alignments would be expected by chance in a database of that size. For a database of 500 million sequences, an E-value of 0.05 is actually quite significant because the chance of observing such an alignment by chance is very low.
Conversely, a hit with a very low E-value can still be biologically meaningless if the alignment covers only a small fraction of the query or database sequence. A 30-amino acid region of high similarity in a 500-amino acid protein may produce a low E-value, but the biological significance depends on whether that region represents a functional domain or a chance similarity.
The E-value also depends on the database size. Searching a large database produces higher E-values for the same alignment score than searching a small database, because there are more opportunities for chance matches. This means that E-values from different searches are not directly comparable unless the database sizes are similar.
Database Contamination
Database contamination is a growing concern in BLAST searches. Contamination can arise from vector sequences in cloned DNA, adapter sequences in high-throughput sequencing, or cross-species contamination in genome assemblies. The most insidious form is cross-contamination, where sequences from one species are incorrectly attributed to another.
A classic example is the presence of human sequences in bacterial genome assemblies, which can occur when human DNA contaminates the bacterial culture or sequencing library. A BLAST search against the bacterial genome will return human hits with high scores, which can be misinterpreted as evidence of horizontal gene transfer. The Transposable Element content of many genomes can also produce spurious hits, as transposable elements are often conserved across species.
To detect contamination, examine the taxonomic distribution of hits. If a search against a bacterial genome returns many hits to human or other mammalian sequences, contamination is likely. The NCBI contamination screening pipeline can be used to identify and remove such sequences from assemblies.
Handling Low Complexity Regions
Low complexity regions are a frequent source of spurious BLAST hits. A query containing a poly-proline tract will produce high-scoring alignments with any protein containing a similar tract, regardless of evolutionary relationship. The default filtering should be left enabled for most searches, but there are situations where filtering is problematic.
For proteins with functionally important low complexity regions, such as the prion protein's octapeptide repeats or transcription factor activation domains, filtering will mask the biologically relevant features. In such cases, the search should be performed with filtering disabled, but the results should be carefully examined for spurious hits.
A useful approach is to perform two searches: one with filtering enabled and one with filtering disabled. The filtered search provides the primary results, while the unfiltered search reveals whether the top hits are mediated by low complexity regions. If the top hits in the unfiltered search are different from those in the filtered search, the low complexity regions are likely driving the alignments.
Practical Summary: Best Practices for BLAST NCBI Search
Step-by-Step Workflow
A successful BLAST search requires careful planning and execution. The following workflow provides a systematic approach:
- Prepare the query sequence. Ensure the sequence is in FASTA format and contains only the sequence of interest. Remove vector sequences, adapters, and low-quality regions. For protein queries, verify that the sequence is in the correct reading frame.
- Select the appropriate BLAST program. Choose blastn for nucleotide-nucleotide comparisons, blastp for protein-protein comparisons, and blastx or tblastn for translated searches. The choice depends on the query type and the database to be searched.
- Choose the database. Select nr for comprehensive protein searches, nt for comprehensive nucleotide searches, RefSeq for curated sequences, or an organism-specific genome database for mapping to a particular species.
- Set the parameters. Adjust the E-value threshold, word size, and scoring matrix based on the expected divergence between the query and the target sequences. Use the default parameters for initial searches and refine as needed.
- Run the search and examine the results. Review the hit table, graphical summary, and alignments. Pay attention to query coverage, identity, and the taxonomic distribution of hits.
- Validate the results. Confirm that the top hits are biologically plausible. Check the annotation of the database sequences and consider whether the hits represent orthologs or paralogs. The Gene Duplication events that generate paralogs can complicate interpretation, as can the Positive and Negative Selection pressures that shape sequence evolution.
- Document the search parameters. Record the BLAST version, database version, and all parameters used. This is essential for reproducibility and for comparing results across searches.
Documenting Your Search
Reproducibility is a cornerstone of scientific research, and BLAST searches are no exception. The NCBI BLAST portal provides a "Search ID" for each search, which can be used to retrieve the results within 24 hours. For permanent documentation, the search parameters and results should be saved.
The command-line BLAST+ suite provides greater control and reproducibility than the web interface. A typical command-line search might be:
blastp -query protein.fasta -db nr -out results.txt -outfmt 6 -evalue 1e-5 -max_target_seqs 100The -outfmt 6 option produces a tabular output that is suitable for downstream analysis, while -max_target_seqs limits the number of hits reported. The command-line tools also allow the use of custom databases, which is essential for comparing sequences against unpublished data.
For large-scale analyses, the results should be stored in a structured format that can be parsed programmatically. The tabular output format includes the query ID, subject ID, percentage identity, alignment length, number of mismatches, number of gap openings, query start and end positions, subject start and end positions, E-value, and bit score. This information is sufficient for most downstream analyses, including phylogenetic reconstruction and Neutral Theory of Molecular Evolution tests.
Frequently Asked Questions
How do I perform a BLAST search on NCBI?
Navigate to the NCBI BLAST portal (https://blast.ncbi.nlm.nih.gov), select the appropriate BLAST program (blastn for nucleotide, blastp for protein), paste your query sequence in FASTA format or enter an accession number, select a database, and click "BLAST." The results page will display the hit table, graphical summary, and alignments. For a more detailed guide, see the Blast Primer Search article.
What does E-value mean in BLAST results?
The E-value (Expectation value) represents the number of alignments with a score at least as high as the observed score that would be expected to occur by chance in a database of the same size. An E-value of 0.001 means that one such alignment would be expected by chance in 1,000 searches of a database of that size. Lower E-values indicate more significant matches. The E-value is not a probability of a false positive; it depends on the database size and the scoring system.
Which BLAST program should I use for a nucleotide query?
For a nucleotide query, the choice depends on the database you want to search. Use blastn to search a nucleotide database, such as nt or a genome database. Use blastx to search a protein database, which is useful for identifying the protein encoded by an unannotated nucleotide sequence. Use tblastx to search a translated nucleotide database, which is computationally expensive and rarely necessary.
How do I choose the right database for BLAST?
The choice of database depends on the biological question. For identifying the function of an unknown protein, search the non-redundant (nr) protein database. For mapping a sequence to a specific genome, use the organism-specific genome database. For curated, well-annotated sequences, use the RefSeq database. For transcriptome analysis, use the EST or TSA databases. The Blast Basic Local Alignment Search Tool article provides additional guidance.
Why is my BLAST search returning no hits?
A BLAST search can return no hits for several reasons. The query may be too divergent from any sequence in the database, the E-value threshold may be too stringent, the database may not contain sequences from the relevant organism, or the query may contain low complexity regions that are filtered out. Try increasing the E-value threshold, reducing the word size, disabling low complexity filtering, or searching a different database.
What is the difference between identity and similarity in BLAST?
Identity is the percentage of aligned positions where the residues are identical between the query and the database sequence. Similarity is a broader concept that includes both identical residues and conservative substitutions, which are residues with similar physicochemical properties that score positively in the substitution matrix. In BLAST output, the "+" symbol in the alignment indicates conservative substitutions. Similarity is always greater than or equal to identity.
How can I improve BLAST sensitivity for distant homologs?
To improve sensitivity for distant homologs, use PSI-BLAST, which constructs a position-specific scoring matrix from an initial search and iteratively searches for additional homologs. Alternatively, reduce the word size, use a more permissive substitution matrix such as BLOSUM45, increase the E-value threshold, or disable low complexity filtering. The Conserved Sequence analysis can help identify the functionally important residues that are most useful for detecting distant homologs.
Key Takeaways
- BLAST is a heuristic local alignment tool that identifies regions of similarity between a query sequence and a database, with statistical significance assessed by E-values derived from extreme value distribution theory.
- The choice of BLAST program (blastn, blastp, blastx, tblastn, tblastx) depends on the query type and the database to be searched, with translated searches essential for identifying protein-coding regions in unannotated nucleotide sequences.
- Database selection is critical: nr provides comprehensive coverage, RefSeq offers curated annotations, and organism-specific genome databases enable precise mapping to a particular species.
- E-values are not probabilities; they represent the expected number of chance alignments in a database of a given size, and they depend on database size and scoring parameters.
- Low complexity filtering is essential for most searches but can mask biologically meaningful features, requiring careful interpretation when disabled.
- PSI-BLAST and PHI-BLAST provide advanced capabilities for detecting distant homologs and motif-containing proteins, but they require careful parameter selection to avoid contamination by false positives.
- Reproducibility requires documenting all search parameters, including BLAST version, database version, E-value threshold, word size, and scoring matrix, with command-line BLAST+ providing the most rigorous control.