File size: 3,483 Bytes
b2ca83f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 | # =============================================================================
# config.yaml — Genome Dataset Pre-processing Configuration
# =============================================================================
# Root directory containing all species folders.
# Expected layout: {data_root}/{species}/{assembly}/{species}_{annotation}_genomic.fna
# {species}_{annotation}.gff3
data_root: /share/kuleshov/emm392/mammal_genomes/
# Where to write the final HuggingFace DatasetDict (saved with save_to_disk).
output_dir: "./output"
# Where to cache processed annotation files
cache_dir: ".cache"
# ---------------------------------------------------------------------------
# Sliding-window parameters
# ---------------------------------------------------------------------------
chunk_size: 12_000 # Length of each sequence chunk in base pairs
stride: 12_000 # Step size between consecutive chunks (256 = 50% overlap for chunk of 512)
# Set stride == chunk_size for non-overlapping chunks
# ---------------------------------------------------------------------------
# CDS flanking regions
# ---------------------------------------------------------------------------
# Extra bases to include BEFORE the CDS start (upstream, on the feature strand)
flank_upstream_bp: 10_000
# Extra bases to include AFTER the CDS end (downstream, on the feature strand)
flank_downstream_bp: 10_000
# max percentage of N's allowed in a given sequence
max_n_perc: 0.25
# ---------------------------------------------------------------------------
# Species selection & validation split
# ---------------------------------------------------------------------------
# Only species listed here are processed — everything else in data_root is
# ignored. Each entry requires:
# - name : matches the directory name exactly under data_root
# - val_chromosome: contig/chromosome ID to hold out for validation.
# Must match the FASTA header exactly (first word after ">").
# Tip: grep "^>" your_file.fna | head to list available IDs.
#
# train species — ALL chromosomes go to train except val_chromosome
# validation species — same rule applies; val_chromosome goes to validation,
# remaining chromosomes go to train
#
# If you want a species to contribute ONLY to train (no val chrom), set
# val_chromosome to null.
species:
train:
- name: "Homo_sapiens"
val_chromosome: "NC_000008.11" # hold out chr8
- name: "Mus_musculus"
val_chromosome: "NC_000070.7" # hold out chr3
- name: "Pan_troglodytes"
val_chromosome: "NC_072404.2" # hold out chr6
# Add more training species here:
# - name: "rattus_norvegicus"
# val_chromosome: "NC_005100.4"
validation:
# Species listed here follow the same rule: val_chromosome → validation,
# all other chromosomes → train. Use this section if you want to keep
# certain species exclusively (or primarily) for evaluation bookkeeping.
# Most users will leave this empty and rely on val_chromosome above.
#
# - name: "danio_rerio"
# val_chromosome: "NC_007112.7"
# ---------------------------------------------------------------------------
# Misc
# ---------------------------------------------------------------------------
shuffle: False # Shuffle the training set after building
seed: 42
|