Default Step Config¶

SpaceTracer/config/default_step_config.yaml stores the default parameter values used by SpaceTracer steps.

This file is for advanced step tuning. It does not replace config.yaml: config.yaml is still the input file passed to spacetracer run --config config.yaml. In the current version, if advanced users want to modify step parameter values, they should edit SpaceTracer/config/default_step_config.yaml directly.

For workflow order and detailed step behavior, see:

• Step-by-step guide
• Single-Step Debug Cookbook
• cluster
• bam_processing
• mpileup
• umi_combine
• cell_num
• prior
• genotyping
• spatial_feature
• mappability_feature
• read_feature
• RNA_feature
• phasing
• feature_filtration / merge_feature
• mutation_prediction
• phylogeny

Complete default file¶

steps:
  cluster:
    ncluster: 8
    plot: true
    method: "SpaGCN"
    init_method: "louvain"
    h5_file_name: "filtered_feature_bc_matrix.h5"
    histology: true
    spot_area: 49
    weight_histology: 1
    distance_threshold: 2
    min_samples: 1
    num_threshold: 30
    percentage: 0.5
    seed: 100
    tol: 5e-3
    lr: 0.05
    max_epochs: 200
    graphst_tool: "louvain"
    radius: 6
    refinement: true

  bam_processing:
    nm_threshold: 5
    mapq_threshold: 255

  mpileup:
    min_depth: 30
    max_depth: 200000
    min_mapq: 0
    min_baseq: 0
    exclude_flag: 0
    enable_split: true
    split_threshold: 100000
    chrom_chunk_size: 20000
    chrM_chunk_size: 5000
    max_cost: 200000

  UMI_combine:
    filter_duplicates: true
    filter_secondary: true
    filter_qcfail: true
    filter_supplementary: true
    min_read_quality: 20

  genotyping:
    alpha: 0.05
    epsQ: 20
    epsAF: 0.003
    mu: 1e-5
    thr_dp: 1000
    pop_vaf: 1e-5
    filter_oneallele: true

  spatial_feature:
    alpha: 0.05
    thr_r2: 0.3
    thr_prob: 0.9
    thr_likelihood: 0.9
    thr_vaf: 0
    plot_supp: false
    fig_size: 5
    method: LDA
    num_directions: 8

  read_feature:
    downsample: true
    downsample_target_depth: 2000
    max_region_size: 20000
    max_variants_per_region: 100
    seed: 42

  RNA_feature:
    min_count_for_germline: 50
    min_prior_for_germline: 0.0001
    default_range_of_gene: 150
    p_threshold: 0.05
    previous_base: 5

  phasing:
    minprior: 0.01
    min_dp: 20
    min_total_dp: 50
    alpha: 0.05
    phasing_pad: 1000
    merge_gap: 200
    max_target: 200000
    seed: 42

  feature_filtration:
    LOW_VAF: True
    HIGH_VAF: True
    LOW_SPOT_NUM: True
    LOW_ALT_SPOT_NUM: True
    ASE: True
    hFDR: True
    imprinted: True
    homopolymer: True
    PON: True
    RNA_editing: True
    ABNORMAL_MISMATCHES: True
    LOW_READ_DIVERSITY: True
    HIGH_MULTIPLE_MAPPIN: True
    WIDE_DISTRIBUTION: True
    NEAR_READ_END: True
    CLUSTER_EVENTS: True
    LOW_MAPQ: True
    LOW_BASEQ: True
    INDEL_PROPORTION: True
    ALT_ALLELE_COUNT: True
    POPULATION_AF: True
    CONTIG: True
    MITOCHONDRIA: True
    CLUSTERED_NOISE(RNA_editing): True
    CLUSTERED_NOISE: True

  mutation_prediction:
    random_seed: 42
    plot: true

cluster¶

Controls automatic spot/domain clustering when no cluster_file is provided in config.yaml.

• ncluster: target number of clusters.
• plot: whether to generate clustering plots.
• method: clustering backend. Supported values include default, SpaGCN, and GraphST.
• init_method: initialization method for clustering, such as louvain.
• h5_file_name: Space Ranger feature matrix file name.
• histology: whether to use histology image information.
• spot_area: expected spot area used by spatial clustering.
• weight_histology: weight for histology information.
• distance_threshold: spatial distance threshold for domain refinement.
• min_samples: minimum samples used in neighborhood/domain filtering.
• num_threshold: minimum spot count threshold for domain filtering.
• percentage: spatial/histology weighting parameter used by the clustering backend.
• seed: random seed for reproducible clustering.
• tol: optimization tolerance.
• lr: learning rate.
• max_epochs: maximum optimization epochs.
• graphst_tool: community detection tool used by GraphST.
• radius: neighborhood radius used by GraphST.
• refinement: whether to refine cluster assignments after initial clustering.

bam_processing¶

Controls BAM-level filtering before pileup and read-feature extraction.

• nm_threshold: maximum allowed edit distance/mismatch tag value.
• mapq_threshold: minimum mapping quality threshold.

mpileup¶

Controls candidate-site pileup generation and chunking.

• min_depth: minimum depth required for a site.
• max_depth: maximum depth allowed for a site.
• min_mapq: minimum read mapping quality used by pileup.
• min_baseq: minimum base quality used by pileup.
• exclude_flag: SAM flag mask for reads excluded from pileup.
• enable_split: whether to split large pileup work into chunks.
• split_threshold: threshold for enabling chunk splitting.
• chrom_chunk_size: chunk size for standard chromosomes.
• chrM_chunk_size: chunk size for mitochondrial chromosome processing.
• max_cost: maximum chunk cost used to control workload size.

UMI_combine¶

Controls read and UMI filtering before combined UMI evidence is produced.

• filter_duplicates: remove duplicate reads.
• filter_secondary: remove secondary alignments.
• filter_qcfail: remove reads marked as QC-failed.
• filter_supplementary: remove supplementary alignments.
• min_read_quality: minimum read quality for UMI evidence.

genotyping¶

Controls genotype inference thresholds.

• alpha: statistical significance threshold.
• epsQ: sequencing error quality parameter.
• epsAF: allele-frequency error parameter.
• mu: prior mutation rate.
• thr_dp: depth threshold.
• pop_vaf: population allele-frequency threshold.
• filter_oneallele: whether to filter one-allele-only evidence.

spatial_feature¶

Controls spatial feature extraction from genotype evidence.

• alpha: significance threshold.
• thr_r2: spatial correlation threshold.
• thr_prob: probability threshold.
• thr_likelihood: likelihood threshold.
• thr_vaf: variant allele frequency threshold.
• plot_supp: whether to generate supplementary plots.
• fig_size: output figure size.
• method: spatial feature method, such as LDA.
• num_directions: number of spatial directions used for directional features.

read_feature¶

Controls read-level feature extraction.

• downsample: whether to downsample high-depth regions.
• downsample_target_depth: target depth after downsampling.
• max_region_size: maximum region size processed for read features.
• max_variants_per_region: maximum number of variants processed per region.
• seed: random seed for reproducible downsampling.

RNA_feature¶

Controls RNA-level feature extraction.

• min_count_for_germline: minimum count used when evaluating germline-like evidence.
• min_prior_for_germline: minimum prior used for germline filtering.
• default_range_of_gene: default gene-flanking range.
• p_threshold: p-value threshold.
• previous_base: number of previous bases considered for sequence-context features.

phasing¶

Controls RNA-informed phasing and cluster-event refinement.

• minprior: minimum prior required for phasing.
• min_dp: minimum depth for a phased site.
• min_total_dp: minimum total depth.
• alpha: significance threshold.
• phasing_pad: padding around target sites for phasing.
• merge_gap: maximum gap for merging nearby phased regions.
• max_target: maximum number of target sites.
• seed: random seed.

feature_filtration¶

Controls which filtration tags are applied during feature merge and filtration.

• LOW_VAF: filter low variant allele frequency calls.
• HIGH_VAF: filter unusually high variant allele frequency calls.
• LOW_SPOT_NUM: filter calls supported by too few spots.
• LOW_ALT_SPOT_NUM: filter calls supported by too few mutant spots.
• ASE: filter allele-specific expression artifacts.
• hFDR: filter calls failing hFDR-related criteria.
• imprinted: filter imprinted-region artifacts.
• homopolymer: filter homopolymer-context artifacts.
• PON: filter panel-of-normals artifacts.
• RNA_editing: filter known RNA-editing sites.
• ABNORMAL_MISMATCHES: filter abnormal mismatch-pattern artifacts.
• LOW_READ_DIVERSITY: filter calls with low read diversity.
• HIGH_MULTIPLE_MAPPIN: filter high multiple-mapping artifacts.
• WIDE_DISTRIBUTION: filter calls with overly broad spatial distribution.
• NEAR_READ_END: filter calls close to read ends.
• CLUSTER_EVENTS: filter cluster-level event artifacts.
• LOW_MAPQ: filter low mapping-quality evidence.
• LOW_BASEQ: filter low base-quality evidence.
• INDEL_PROPORTION: filter calls with high nearby indel proportion.
• ALT_ALLELE_COUNT: filter calls with insufficient alternate allele count.
• POPULATION_AF: filter common population variants.
• CONTIG: filter calls on configured contig chromosomes.
• MITOCHONDRIA: filter mitochondrial calls.
• CLUSTERED_NOISE(RNA_editing): filter clustered A-to-G/T-to-C noise patterns consistent with RNA-editing-like artifacts.
• CLUSTERED_NOISE: filter other clustered technical-noise patterns.

mutation_prediction¶

Controls final model-based mutation prediction behavior.

• random_seed: random seed for reproducible prediction-related operations.
• plot: whether to generate mutation prediction plots.