In this project, I am designing machine learning models that predict ecological niches based on genomic data and identifying conserved ESKAPEE genes related to pathogenicity based on niche localization. Based on proportion of genomes passing quality metrics, I am starting with A. Baumannii genomes
- Processed genomic metadata retrieved from NCBI GenBank and assessed genome quality metrics, including CheckM contamination, completeness, number of contigs, and N50 contig size
- Completed feature creation by categorizing niches based on a combination of host disease and isolation source metadata
- Made supplementary plots of genome quality and niche proportions
- Plot of Clonal Clusters Showing Proportion of Niches
- Sankey Plot of Genomovars Mapped to Niches
- QA_11_7.rmd to get the AB txt file
# Used genbank metadata to categorize samples into niches and did quality filtering (completion & contamination)
- To download genomic info from a txt file
conda activate ncbi-datasets
datasets download genome accession --inputfile ~/scratch/AB/ABgenomesToDownload_new.txt --dehydrated --filename ~/scratch/AB/AB.zip
unzip ~/scratch/AB/AB.zip -d ~/scratch/AB/AB_dataset
datasets rehydrate --directory ~/scratch/AB/AB_dataset/
- Mash reference list: Create a text file with the filenames of all genomes in directory called reference_list.txt
./reference_list.sh
- Mash query list: Create batch text files (3,925 genomes --> 27 genomes per batch file)
~/scratch_2/splitting_batches.sh <input_file> <num_files>
~/scratch_2/splitting_batches.sh ~/scratch_2/AB/reference_list.txt 146
- Pairwise comparisons with Mash
sbatch mash_try_AB_sketch.sh
./submit_jobs.sh &> submit_jobs.log &
- Clustering with custom script
# combine all files in mash_output into one tsv file
cat $(ls ~/scratch_2/AB/mash_output/) > ~/scratch_2/AB_allMash.tsv
bash clusterMash.sh # run clustering script
- Visualizing the clusters with sankey and proportion plots
# run clusterMash2.ipynb
- Running bakta
nohup ./submit_jobs.sh &> submit_jobs.log &