At the moment the algorithm is temporarly not working due to some changes in the baseline infrastucture for improving memory management.
- Python >= 3.9
- Scipy
- Numpy
- Numba
- Matplotlib
- Inspyred
- Sklearn
- Biopython
This algorithm, instead of the usual De Bujin graph used in the majority of similar methods to performe de novo assembly, builds an Overlap-Layout Consensus (OLC) graph starting from the reads given in input. At this data the only input format available is fastq, however the file passed could be, not only deriving from illumina sequencing platform, but also from both nanopore (ONT) and hifiam (Pacbio) platforms.
After this first step, the graph will be simplified eraising those branches and links which retrive poor overlap between the reads. All this part is done through statistical analysis, not taking in consideration the sequences of the reads or the phred score of them. To assert the presence of contigs (gaps in the coverage), the algorithm has been implemented with a hierarchical clustering of the remaining weighted edges (embedding of the graph will be implemented to retrive further information and statistical analysis). The clustering is performed uscing the weights of the edges as distances between each point (node/read in the graph). This part of clustering is done not only in order to divide the problem in smaller ones, to reduce the demand in space in the RAM usage, but also to increase the efficiency of the ANT colony system downstream.
Aftermath each cluster is given as input to the ant colony system, which resolve the optimum problem of finding the best path (best assembly) and return a candidate solution. The path has to be now converted into sequence, in doing so the algorithm builds a consensus matrix. The final matrix evaluate and return the most probable base at each position, taken in consideration the multiple overlap created by the algorithm and each base phred score. Possibles SNPs or variants will be stored in a separate file.
Some of the features are not already present, however they will be implemented sonn or are currently in the developing process.
usage: ants_assembly.py [-h] [-i INPUT [INPUT ...]] [-o OUTPUT_DIRECTORY] [--test [TEST]]
[-p POPULATION_SIZE] [-e EVAPORATION_RATE] [-r LEARNING_RATE]
[-v VERBOSE] [-cpus CPU_CORES] [-g MAX_GENERATION]
[-L IPOTHETICAL_LENGTH]
options:
-h, --help show this help message and exit
-i INPUT [INPUT ...], --input INPUT [INPUT ...]
The input must be a fasta or fastq file
-o OUTPUT_DIRECTORY, --output_directory OUTPUT_DIRECTORY
Directory of output
--test [TEST] This is for testing
-p POPULATION_SIZE, --population_size POPULATION_SIZE
-e EVAPORATION_RATE, --evaporation_rate EVAPORATION_RATE
Internal parameter of the ant colony system
-r LEARNING_RATE, --learning_rate LEARNING_RATE
Internal parameter for the ant colony system
-v VERBOSE, --verbose VERBOSE
Prints and return more information on how the process is developing
-cpus CPU_CORES, --cpu_cores CPU_CORES
Number of cpu to use; default = 2
-g MAX_GENERATION, --max_generation MAX_GENERATION
Number of iterations/generatios of the ant colony algorithm
-L IPOTHETICAL_LENGTH, --ipothetical_length IPOTHETICAL_LENGTH
For a better reconstruction of the genome an ipotetical lenght of the sequence to rebuild is
fondamental for retriving good resultsUp to date the algorithm function only with option --test or with fastq file of small dimensions (order of some Mb in total).
- Use of the graph embedding to analyse contigs
- Implementation of hierarchical clustering