blast_samples_extended/
========================

Three further tests extending the chr1 BIG-cluster BLAST finding (see
blast_samples/README.txt), each answering a different question.


FILES
-----

consensus_samples_extended.fasta
    24 sequences. Representative sequences: up to 6 each of BIG (>=1000),
    MEDIUM (100-300), SMALL (5-15), SINGLETON (1), spread across chr1-chr6
    for broader chromosome coverage. Headers labeled
    CATEGORY_block_sizeN_originalID.

full_big.fasta / full_big_members.tsv
    3823 sequences. BIG cluster, locus chr1:100330740-100565735 (a
    different chr1 block than the original chr1:10023878-10175888 test).

full_medium.fasta / full_medium_members.tsv
    285 sequences. MEDIUM cluster, SAME locus as full_big.fasta above
    (chr1:100330740-100565735) -- deliberately chosen so BIG/MEDIUM/SMALL
    come from one fixed haploblock, isolating cluster size as the only
    variable.

full_small.fasta / full_small_members.tsv
    8 sequences. SMALL cluster, same locus as above.

full_telo_max.fasta / full_telo_max_members.tsv
    3 sequences. Largest cluster found in a telomeric, near-maximum-entropy
    locus, chr20:1-598702 (entropy=12.3125, dominance=0.00059 -- essentially
    every haplotype in this block is distinct).

full_telo_min.fasta / full_telo_min_members.tsv
    1 sequence. Smallest (singleton) cluster in the same chr20 locus.
    598,676 bp, representative NA20534_chr20_region_1-598702_hap1.

representative_candidates.tsv
    Selection table for the 24 general BIG/MEDIUM/SMALL/SINGLE candidates.


PURPOSE OF EACH PART OF THIS BATCH
------------------------------------

MEDIUM category, same-locus design:
    Does the representative-selection failure scale continuously with
    cluster size, or switch on above some threshold? Fixing the locus
    across BIG/MEDIUM/SMALL isolates size as the only variable.

Broader BIG/SMALL/SINGLE coverage (chr1-chr6):
    Does the original finding hold up across more, and more varied, loci?

Telomeric high-entropy locus (chr20:1-598702):
    Does the failure mode require a dominated cluster (one common
    haplotype background) to exist at all? In this locus the largest
    cluster found in a genome-wide search was only 3 members -- there is
    no "big" cluster here for the representative-drift mechanism to act
    on, making this a structural contrast case rather than a smaller
    replicate of the chr1 result.


GENERATION
----------

Produced by prepare_extended_datasets.sh from block_stats.tsv,
clusters_summary.tsv, and <chr>/clusters/<block>_{rep_seq,all_seqs}.fasta.
Re-extracted with reextract_with_fixed_dedup.sh on 2026-09-05 to correct
the duplicate-representative bug described above.