Data for "A Hybrid Pipeline for Feature Reduction, and Ordinal Classification to Predict Antimicrobial Resistance from Genetic Profiles"

Notation Rules for Variable Interpretation We provide below the notation system used to encode gene-level features in the dataset: {P,N}_{A,P}#_{R,P}#_{A,P}X#Y The notation is divided by underscores (_). The symbol # represents a number composed of at least one digit. Curly braces {} indicate allowed options. For example, {P,N} means either P or N must appear in that position. Each entry consists of three or four components. 1. Genomic Location Indicates whether the gene is located on a plasmid. P – Located on a plasmid N – Not located on a plasmid 2. Gene Function Indicates the functional annotation source. A – Function annotated in CARD P – Function identified by Prokka – Function identifier When the function type is A, # corresponds to a CARD ARO identifier. When the function type is P, # corresponds to an autogenerated identifier that maps to a standardized functional annotation index provided in a companion file. For example: P0 → hypothetical protein P1 → tRNA U34 carboxymethyltransferase P2 → Cytosol non-specific dipeptidase P3 → Proline/betaine transporter P4 → ABC transporter glutamine-binding protein GlnH P5 → Hippurate hydrolase ... The complete mapping between Prokka-derived identifiers (P#) and their functional descriptions is provided in the accompanying file: Prokka_Function_Reference_Table.txt 3. Molecule Type and Family Indicates the type of gene product. R – The gene produces an RNA P – The gene produces a protein – Autogenerated identifier, unique within each function category 4. Optional Mutation Annotation Indicates the presence of a mutation. A – Mutation annotated in CARD P – Mutation detected through gene family alignment X – Reference nucleotide or amino acid For P-type mutations, the reference corresponds to the most common nucleotide or amino acid within the alignment (or the first in alphabetical order in case of ties). For A-type mutations, the reference corresponds to a CARD gene from a susceptible strain. If X = "-", this indicates an insertion relative to the reference. – Position of the mutation in the RNA or protein sequence Y – New nucleotide or amino acid If Y = "-", this indicates a deletion relative to the reference. Example feature name P_A3008823_P0 corresponds to a gene located on a plasmid that produces a protein with CARD ARO identifier 3008823 (PC1 protein). Annotation Priority Rules CARD annotations (functions and mutations) take precedence over Prokka and alignment-derived annotations, respectively. CARD annotations replace Prokka or alignment annotations when all of the following conditions are met: Same contig Same strand Same gene type (RNA or protein) Overlap greater than 95% Counting Rules Gene counts and mutation counts are reported in separate columns. Each column represents counts from a single gene family. A gene family may appear across multiple columns. Genes annotated as having unknown function are not guaranteed to be grouped consistently. File Descriptions pangenome.npz. A SciPy Compressed Sparse Row (CSR) matrix containing counts with ~29K rows and ~7M columns (features), as uint8 data type (to save space and memory). pangenome.features.txt. Stores the names of the 7M features (columns) of pangenome.npz, one line per column. pangenome.json. Maps Prokka function identifiers (starting with P, see section 2 of the notation rules above) to their function names. For CARD functions, you can look up the functions on the CARD website. pangenome.metadata.csv. Stores the metadata of each of the ~29K rows of pangenome.npz. Specifically, the following columns: accession. Unique identifier of the genome (taken from NCBI, ENA, BV-BRC). dataset. List of CAMDA dataset it appears in (2025 or 2026, training, balancing or testing sets) biosample. The BioSample identifier this genome belongs to. antibiotic. Antibiotic... Zenodo record 19824655, published 2026-04-27, 4 file(s) totalling 1966.8 MB (2 tabular, 1 other, 1 archive).

Category
Drug Discovery
Provider
Anton Pashkov
License
CC-BY-4.0
DOI
10.5281/zenodo.19824655
Hosted on
Zenodo
Access
Free and public. No account with the data provider is needed.
Source
https://doi.org/10.5281/zenodo.19824655
Data structure and loading code

Fields

accessionstring
datasetstring
biosamplestring
antibioticstring
speciesstring
micfloat
phenotypestring
gtdb_speciesstring
gtdb_genusstring
gtdb_anifloat
adjusted_micfloat
adjusted_phenotypestring

Files

Example file: https://zenodo.org/api/records/19824655/files/pangenome.metadata.csv/content

Path pattern: Flat Zenodo record: every file hangs directly off the record, no directory tree.

Load the data

import pandas as pd

# pangenome.metadata.csv (4.6 MB) streamed straight from Zenodo.
df = pd.read_csv("https://zenodo.org/api/records/19824655/files/pangenome.metadata.csv/content", sep=",", low_memory=False)
print(df.shape)
df.head()

Written for a Scigantic notebook, where /mnt/archive is the mounted dataset.

Analyze this dataset without downloading it

On Scigantic, Data for "A Hybrid Pipeline for Feature Reduction, and Ordinal Classification to Predict Antimicrobial Resistance from Genetic Profiles" opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

Related Drug Discovery datasets