mirror of
https://github.com/metabarcoding/obitools4.git
synced 2026-04-30 12:00:39 +00:00
37 lines
1.8 KiB
Markdown
37 lines
1.8 KiB
Markdown
|
# Semantic Description of `obikmer` Package
|
|||
|
|
|
|||
|
|
This Go package provides utilities for **k-mer (specifically 4-mer) counting and comparison** of biological sequences.
|
|||
|
|
|
|||
|
|
## Core Functionalities
|
|||
|
|
|
|||
|
|
1. **`Count4Mer(seq, buffer, counts)`**
|
|||
|
|
Counts occurrences of all possible 16-mer (4-nucleotide) subsequences in a `BioSequence`.
|
|||
|
|
- Encodes each 4-mer into an integer (0–255) using `Encode4mer`.
|
|||
|
|
- Populates a fixed-size `[256]uint16` table (`Table4mer`) with counts.
|
|||
|
|
- Reuses or allocates the `counts` buffer as needed.
|
|||
|
|
|
|||
|
|
2. **`Common4Mer(count1, count2)`**
|
|||
|
|
Computes the *intersection* of two 4-mer frequency profiles: sum over all k-mers of `min(count1[k], count2[k])`.
|
|||
|
|
Used to measure shared content between sequences.
|
|||
|
|
|
|||
|
|
3. **`Sum4Mer(count)`**
|
|||
|
|
Returns the total number of 4-mers in a profile (i.e., sum over all entries).
|
|||
|
|
|
|||
|
|
## Distance & Similarity Bounds
|
|||
|
|
|
|||
|
|
4. **`LCS4MerBounds(count1, count2)`**
|
|||
|
|
Estimates bounds for the *Longest Common Subsequence* (LCS) length between two sequences based on 4-mer profiles:
|
|||
|
|
- **Lower bound**: `common_kmers + (3 if common > 0 else 0)`
|
|||
|
|
- **Upper bound**: `min(total1, total2) + 3 − ceil((min_total – common)/4)`
|
|||
|
|
Leverages the fact that overlapping k-mers constrain possible alignments.
|
|||
|
|
|
|||
|
|
5. **`Error4MerBounds(count1, count2)`**
|
|||
|
|
Estimates bounds for *alignment errors* (e.g., mismatches + indels):
|
|||
|
|
- **Upper bound**: `max_total − common_kmers + 2 * floor((common_kmers + 5)/8)`
|
|||
|
|
- **Lower bound**: `ceil(upper_bound / 4)`
|
|||
|
|
Provides fast, approximate error estimates without full alignment.
|
|||
|
|
|
|||
|
|
## Use Case
|
|||
|
|
|
|||
|
|
Designed for **high-performance comparison of NGS reads** (e.g., in metabarcoding), where exact alignment is too costly, and k-mer-based heuristics enable scalable similarity estimation.
|