mirror of
https://github.com/metabarcoding/obitools4.git
synced 2026-04-30 12:00:39 +00:00
Eric Coissac
8c7017a99d
- Update obioptions.Version from "Release 4.4.29" to "/v/ Release v5" - Update version.txt from 4.29 → .30 (automated by Makefile)
45 lines
2.6 KiB
Markdown
45 lines
2.6 KiB
Markdown
# Semantic Description of the `obikmer` Package
|
||
|
||
This Go package implements a **De Bruijn graph** for efficient k-mer manipulation and sequence assembly, primarily used in bioinformatics (e.g., metagenomic read error correction or consensus building).
|
||
|
||
### Core Functionalities
|
||
|
||
- **K-mer Encoding**: K-mers are encoded as `uint64` using 2 bits per nucleotide (A=0, C=1, G=2, T=3), supporting IUPAC ambiguity codes via the `iupac` map.
|
||
- **Reverse Complement Handling**: The `revcompnuc` table enables nucleotide-wise reverse complementation.
|
||
- **Graph Construction**: The `DeBruijnGraph` struct maintains a map from k-mer hashes to integer weights (e.g., observed counts), with helper masks for bit manipulation (`kmermask`, `prevc/g/t`).
|
||
|
||
### Graph Operations
|
||
|
||
- **Node Queries**:
|
||
- `Previouses()` / `Nexts()`: Return predecessor/successor k-mers in the graph.
|
||
- `MaxNext()` / `MaxHead()`: Find neighbors or heads (sources) with maximum weight.
|
||
- **Path Exploration**:
|
||
- `MaxPath()`: Greedily traces the highest-weight path from a head.
|
||
- `LongestPath()`: Explores all heads to find the path with maximum cumulative weight (optionally bounded in length).
|
||
- `HaviestPath()`: Uses Dijkstra-like priority queue to find the *heaviest* (sum-weight) path, with cycle detection via DFS (`HasCycle()`).
|
||
|
||
### Consensus & Filtering
|
||
|
||
- **Consensus Generation**:
|
||
- `BestConsensus()` returns a sequence from the greedy max-weight path.
|
||
- `LongestConsensus(id, min_cov)` trims low-coverage ends using a coverage threshold (mode-based).
|
||
- **Weight Statistics**:
|
||
- `MaxWeight()`, `WeightMean()`, `WeightMode()` provide distribution summaries.
|
||
- `FilterMinWeight(min)` removes low-count nodes.
|
||
- **Decoding**:
|
||
- `DecodeNode()` converts a k-mer index to its DNA string.
|
||
- `DecodePath()` reconstructs the full consensus from a path.
|
||
|
||
### I/O & Diagnostics
|
||
|
||
- **GML Export**: `WriteGml()` outputs a directed graph in Graph Modelling Language (for visualization), with edge thickness and labels reflecting weights.
|
||
- **Hamming Distance**: `HammingDistance()` computes edit distance between two encoded k-mers using bit operations.
|
||
- **Sequence Insertion**: `Push()` adds a biosequence (with count weight) to the graph, expanding all IUPAC variants recursively.
|
||
|
||
### Dependencies & Design
|
||
|
||
- Leverages `obiseq` for sequence representation and `logrus`/`slices`/`heap` from Go’s stdlib.
|
||
- Designed for scalability and speed, using bit-level operations to minimize memory footprint.
|
||
|
||
Overall: a robust k-mer graph engine for *de novo* assembly, error correction, and consensus recovery in high-throughput sequencing data.
|