Exploring the evolutionary conservation and sequence grammar of the triplicated LEPQQYEV repeat in the N-terminal intrinsically disordered region (IDR) of Arabidopsis thaliana FLK.
FLK (Flowering Locus KH domain, UniProt Q9SR13, locus AT3G04610) is a 577-aa KH-domain RNA-binding protein in the autonomous flowering pathway that acts upstream of FLC, and has since been shown to act as an mRNA m6A reader affecting FLC transcript stability/splicing and floral transition (Amara et al. 2023).
Mockler et al. (2004) already reported that the FLK N-terminal region (before the first KH domain) is glutamine-rich and contains three perfect copies of the 8-aa repeat LEPQQYEV — this is not a new motif discovery. They also noted the repeat is not strictly conserved in their putative rice ortholog.
This repository asks a narrower, new question: is the repeat's precise
sequence order unusual relative to the local amino-acid composition of the
region it sits in, and how does the repeat change among 22 curated FLK
homologs/ortholog candidates — each with UniProt provenance and
KH-domain-architecture QC, spanning close Arabidopsis relatives to distant
eudicots? No reciprocal-best-hit, synteny, or phylogenetic reconstruction
was performed, so "curated homolog" here is not a claim of formal
orthology (see docs/METHODS.md for the orthology-confidence tiers used
instead). This repository does not claim FLK undergoes liquid-liquid
phase separation (LLPS), that LEPQQYEV drives LLPS, or that the repeat is
an experimentally proven functional motif — see docs/METHODS.md for the
interpretation rules this project holds itself to.
- Mockler et al., PNAS (2004). 10.1073/pnas.0404552101
- Lim et al., Plant Cell (2004). 10.1105/tpc.019331
- Amara et al., Molecular Plant (2023). 10.1016/j.molp.2023.04.005
Full literature context is in docs/LITERATURE_CONTEXT.md, including the
2004 rice observation: Lim et al.'s cited accession (AAL31692) was
independently confirmed identical to the modern UniProt record used here
(Q10GH8), though biological orthology itself was not experimentally
established by either the original paper or this project.
data/ sequence, homolog, and provenance source files
src/ flk_repeat/ analysis package
scripts/ analysis + figure-generation scripts, and reproduce_all.py
results/ generated TSV tables
figures/ generated PNG figures
docs/ methods and literature-context notes
tests/ pytest suite
- Sequence provenance — validate the local Q9SR13 record (accession, length, SHA256) against a stored provenance record.
- FLK sequence architecture — locate the three LEPQQYEV repeats and map KH domains / disorder onto the sequence, across three a priori N-terminal analysis windows (primary 1-150, sensitivity 1-107 / 1-182).
- Homolog/ortholog-candidate conservation (primary analysis) — exact
identity, BLOSUM62-based conservative-substitution similarity, and local
E/Q/Y composition of each repeat across 22 curated candidates (UniProtKB
provenance, discovered via automated homology naming + KH-domain-
architecture QC, never a bare BLAST hit), spanning close Arabidopsis
relatives to distant eudicots and a literature-linked rice candidate.
Every excluded candidate is recorded with a reason in
data/provenance/homolog_provenance.tsv. Exact repeat identity is high among the closest Arabidopsis relatives and generally lower among the more distant sampled homologs — an independent full-length MSA sensitivity check (figures/figS1_alignment_validation.png,docs/METHODS.md) confirms this for the closest relatives but finds the specific per-position pattern alignment-sensitive for several more distant Brassicaceae. - Composition-preserving null models (secondary analysis) — is the observed triplication of an exact 8-mer unusual given the amino acid composition of the N-terminal region? A global null (every residue shuffled) and a segmental null (shuffled only within 25-aa blocks, checked at 5 block-phase offsets for robustness) are each run for 1,000,000 composition-preserving permutations (each empirical p-value reported with an exact Clopper-Pearson 95% CI), alongside an exact analytical probability for the motif itself.
pip install -r requirements.txt
python scripts/reproduce_all.py
python -m pytest tests/reproduce_all.py validates source data, regenerates every table in
results/ and every figure in figures/, and runs consistency checks; it
exits non-zero on failure. It requires no file outside this repository and
no network access.