Taykhoom/siraj-saturation-mutagenesis

Siraj saturation mutagenesis

Base-resolution saturation-mutagenesis measurements from Siraj et al., joined to exact assayed oligo sequences. The release contains the same 227,046 construct IDs in HepG2 and K562, for 454,092 total rows across 159 parent contexts.

The public table contains 39 selected columns. The complete 43-column source table remains reproducible with --full in the linked processing repository.

The table supports motif-grammar, compensatory-edit, local dependency, and model-localization analyses. Cell-type names are physical access partitions, not machine-learning train/test assignments. Group mutation families by parent_context_id (or at least background_id) for held-out evaluation. The source release defines no canonical machine-learning split and no official evaluation metric.

Public log2_skew is the source log2Skew and satisfies log2_skew = activity - baseline_activity.

Oligo sequence convention

sequence contains the designed assay insert only, with no synthesis adapters. A full synthesized oligo was:

ACTGGCCGCTTGACG + stored_sequence + CACTGCGGCTCCTGC

If starting from a full synthesis sequence, trim sequence[15:-15].

sequence_geometry_valid checks every applied mutation. For baseline mutation_code == "m0" rows, mutation_position and mutation_base are focal-variant metadata rather than an applied mutation.

Sequences have source-native lengths 189, 192, 197, 199, or 200 bases. All 9,472 rows marked as indel designs are retained without resizing.

Columns

Column	Meaning
source_index	Zero-based data-row index in satmut.txt.gz.
pair_id	Unique assayed construct identifier, including background and mutation code.
sequence	Exact official assayed insert for this construct.
log2_skew, log2_skew_se, log2_skew_fdr	Parent-relative activity effect, standard error, and false-discovery-rate value.
cell_line	HEPG2 or K562.
parent_context_id	Parent regulatory element or variant-pair context shared by a saturation family.
background_id	Exact allele/window background on which the mutation was introduced.
allele	Background allele state; pair backgrounds can contain two characters.
variant1, variant2	First and optional second source variant identifiers.
center_variant, window	Centered source variant and oligo-window placement.
chrom, pos	Source-intended hg19/GRCh37 genomic coordinate. The upstream release documents that assembly intent, but cohort-wide coordinates were not independently revalidated locally; pos is not an oligo index.
ref, alt, variant2_ref, variant2_alt	Reference/alternate alleles for the first and optional second source variant.
mutation_position	One-based edit position inside sequence; this is distinct from source pos. On m0, it is focal-variant metadata rather than an applied edit.
mutation_base	Base present at mutation_position for applied mutations; focal reference metadata on m0.
is_haplotype	Whether the assayed background contains a variant pair.
mutation_code	Mutation code; m0 denotes the unmutated baseline.
is_indel	Whether the released row is marked as an insertion/deletion design.
dna_mean	Mean plasmid/DNA abundance for the construct.
activity, activity_se, activity_padj	Absolute reporter activity, standard error, and adjusted significance.
variant1_is_emvar, variant2_is_emvar	Whether either source variant was an expression-modulating variant in the main MPRA.
is_variant1, is_variant2	Whether the saturation edit falls at source variant 1 or variant 2.
emvar_variant	Source emVar identifier when the edit is at an emVar site.
baseline_activity, baseline_activity_se	Unmutated-parent activity baseline and its standard error.
posterior_activity, posterior_log2_skew	Empirical-Bayes shrunken absolute and parent-relative effects.
sequence_geometry_valid	True when an applied mutation matches mutation_position/mutation_base; m0 rows are valid baselines by definition.

The complete --full output retains every processed source column, including the redundant sequence length. It is not uploaded to this Hub repository.

Processing and verification

The complete downloader, sequence join, mutation-geometry checks, source hashes, closed-world manifest verifier, and optional deterministic same-implementation reconstruction are available at:

https://github.com/TaykhoomDalal/Siraj-Processing/tree/main/Saturation-Mutagenesis

Stable repository-wide processing documentation: https://github.com/TaykhoomDalal/Siraj-Processing/blob/main/README.md

Citation

Siraj, L. et al. Functional dissection of complex trait variants at single-nucleotide resolution. Nature (2026). https://doi.org/10.1038/s41586-026-10121-6

License

The generated Parquets derive from the CC BY-NC 4.0 Zenodo measurement and sequence tables. The processing implementation is MIT; consulted upstream GPL-3.0 analysis code is not copied into the dataset. Nature supplementary material has separate CC BY 4.0 terms and is documented component-by-component in the processing repository's NOTICE.md.