BLAST E-value Calculator

Calculate BLAST E-value from bit score and database size to assess sequence similarity significance.

Bit score S' (from BLAST output) (bits)

Database size m (number of letters) (letters)

Human genome: 3×10⁹

Query length n (aa or bp)

Raw score S (optional)

Please check your inputs and try again.

Why E-value Depends on Database Size

The E-value is the number of hits with at least that score expected by chance in a database of that size. It therefore scales directly with how much sequence you searched.

E = m × n × 2^−S′

The consequence catches people out: the same alignment gets a worse E-value in a larger database. Searching a bigger space means more chances for a random match of that quality, so significance falls even though the alignment is unchanged.

Database	Size	E-value for a 40-bit hit
Small custom set	10⁶ letters	~0.0003
SwissProt	2 × 10⁸	~0.05
nr	10¹¹	~27

The bit score, by contrast, is normalised for scoring matrix and database size, which is why bit scores are comparable across searches and E-values are not. Reporting a bit score alongside the E-value lets others interpret the result independently of when the search was run — databases grow, so an E-value from five years ago is not reproducible today.

Interpreting the Numbers

E-value	Interpretation
< 10⁻⁵⁰	Near-identical or highly conserved homolog
10⁻⁵⁰ to 10⁻⁵	Confident homology
10⁻⁵ to 0.01	Probably homologous — verify
0.01 to 1	Suspicious — check alignment manually
> 1	Expected by chance

Each bit doubles the significance, so a 10-bit difference is roughly a thousandfold change in E-value. That is why bit score differences that look small are often decisive.

Why Low E-value Does Not Prove Homology

Low-complexity regions — poly-glutamine tracts, coiled coils, proline-rich segments — produce highly significant E-values between unrelated proteins simply because compositionally biased sequences match each other readily. Filtering is applied by default for this reason.

Short high-identity matches are the other trap. A 15-residue perfect match may reflect a shared motif rather than common ancestry, which is why alignment length and coverage matter as much as the E-value itself.

Worked Examples

Example 1: Strong hit: S=120 bits, m=3e9, n=300

E=3e9×300×2^(-120)

Result: E=6.8×10⁻²⁵ — extremely significant

Clear homolog — same function very likely

Example 2: Weak hit: S=40 bits, m=3e9, n=300

E=3e9×300×2^(-40)

Result: E=8.2×10⁻¹ — not significant

Could be chance alignment

Example 3: Same hit, different databases

40-bit alignment in SwissProt versus nr

Result: E ≈ 0.05 versus ≈ 27

The alignment is identical. Only the search space changed, and with it the number of chance matches expected at that score.

Example 4: Why bit differences matter

Two hits differing by 10 bits

Result: About a thousandfold difference in E-value

Each bit doubles significance, so 2¹⁰ is roughly 1,000. Small-looking score gaps are often decisive.

Common Mistakes

⚠️

Comparing E-values across different databases

E-value scales with database size, so the same alignment scores worse in a larger database. Use bit scores for comparison.

⚠️

Treating a low E-value as proof of homology

Low-complexity regions produce significant scores between unrelated proteins. Check that the alignment covers a meaningful length and is not compositionally biased.

⚠️

Ignoring alignment coverage

A short perfect match may indicate a shared motif rather than shared ancestry. Coverage and length matter alongside significance.

⚠️

Quoting an E-value without the database and date

Databases grow continuously, so an E-value is not reproducible later. Bit score and database version make the result interpretable.

Frequently Asked Questions

Bit score vs raw score?

Raw score S depends on substitution matrix (BLOSUM62) and gap penalties. Bit score S'=(λS-lnK)/ln2 normalizes for comparison. Every 1 bit of score doubles the number of hits. Higher bit score = more significant.

Choosing database?

Larger database (nr vs SwissProt): same bit score gives higher E-value. Use SwissProt (reviewed, smaller) for sensitive searches. Use nr for comprehensive coverage. RefSeq for specific organisms. E-value scales with m×n.

What does the E-value mean?

The number of alignments with at least that score expected by chance in a database of that size. Lower means more significant.

Why does database size change the E-value?

Because a larger search space gives more opportunities for a chance match of that quality. The alignment itself is unchanged.

Why report bit scores?

They are normalised for scoring matrix and database size, so they remain comparable across searches and over time as databases grow.

Can a low E-value be misleading?

Yes. Low-complexity regions match readily between unrelated sequences, and short high-identity matches may reflect shared motifs rather than common ancestry.