Dataset analysis: `paper`

Run id 20260515-132223_data_paper_seed0 · 2026-05-18T11:46:57+00:00 · championsnet

TL;DR

3,100,154 rows / 3,100,154 unique sequences (0 duplicates).
Mean labels per row: 1.70 / 8. Rows with all 8 labels: 21,556. Rows with zero labels: 0.
WT length rate: 100.00%; canonical AA rate: 100.00%; N-glyco motif rate: 15.17%.
Mean edit distance to WT: 6.21 (median 6, p99 16).
Overall per-position entropy: 0.19 bits; 13/232 positions above 1 bit.

Basic

rows	3,100,154
columns	80
unique_seq_hash	3,100,154
duplicate_rows	0

source_dataset distribution

source	rows	fraction
`paper`	3,100,154	100.00%

label_source distribution

label_source	rows	fraction
`sort_label`	3,100,154	100.00%

Label coverage

Labels per row

labels per row	rows	fraction	distribution
1	2,055,123	66.29%	████████████████████████████████
2	543,388	17.53%	████████░░░░░░░░░░░░░░░░░░░░░░░░
3	224,292	7.23%	███░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
4	116,118	3.75%	█░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
5	66,314	2.14%	█░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
6	41,490	1.34%	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
7	31,873	1.03%	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
8	21,556	0.70%	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

Per-receptor balance

Receptor	binders	non-binders	missing	missing %	bind % (of called)
`R1`	282,856	468,712	2,348,586	75.76%	37.64%
`FcRn6`	286,303	429,758	2,384,093	76.90%	39.98%
`R2AH`	42,020	501,247	2,556,887	82.48%	7.73%
`R2AR`	64,880	600,312	2,434,962	78.54%	9.75%
`R2B`	62,041	607,149	2,430,964	78.41%	9.27%
`R3AF`	50,534	576,958	2,472,662	79.76%	8.05%
`R3AV`	62,755	567,196	2,470,203	79.68%	9.96%
`R3B`	32,788	619,807	2,447,559	78.95%	5.02%

Profile strings

3,074 unique profile strings; 21,556 rows (0.70%) have all 8 receptors labelled.

Top 20 most common profiles

profile	rows	fraction
`---0----`	262,759	8.48%
`----0---`	253,448	8.18%
`-------0`	239,231	7.72%
`0-------`	229,716	7.41%
`------0-`	226,730	7.31%
`-----0--`	223,284	7.20%
`--0-----`	202,668	6.54%
`-0------`	169,363	5.46%
`-1------`	102,227	3.30%
`1-------`	81,823	2.64%
`---00---`	18,578	0.60%
`----0--0`	18,449	0.60%
`-----0-0`	18,283	0.59%
`----00--`	17,930	0.58%
`----0-0-`	17,868	0.58%
`------00`	17,279	0.56%
`---0---0`	16,749	0.54%
`---0-0--`	16,346	0.53%
`---0--0-`	16,161	0.52%
`-------1`	16,098	0.52%

Sequence features

wt_length_rate	1.00
canonical_aa_rate	1.00
n_glyco_motif_rate	0.1517
in_variable_region_only_rate	1.0000

Edit distance to WT

mean = 6.21, std = 2.52, median = 6, p90 = 8, p99 = 16, max = 58.

edit distance	rows	histogram
`0`	0	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
`1`	1,498	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
`2`	100,247	█░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
`3-4`	382,882	█████░░░░░░░░░░░░░░░░░░░░░░░░░
`5-7`	2,128,905	██████████████████████████████
`8-9`	254,420	███░░░░░░░░░░░░░░░░░░░░░░░░░░░
`10-14`	184,445	██░░░░░░░░░░░░░░░░░░░░░░░░░░░░
`15-19`	34,782	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
`20-29`	12,110	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
`30-49`	862	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
`>=50`	3	░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

Mean edit distance, by per-receptor label

Receptor	binders	non-binders	missing
`R1`	mean=5.46, median=6 (n=282,856)	mean=6.09, median=6 (n=468,712)	mean=6.33, median=6 (n=2,348,586)
`FcRn6`	mean=4.89, median=5 (n=286,303)	mean=6.17, median=6 (n=429,758)	mean=6.38, median=6 (n=2,384,093)
`R2AH`	mean=3.23, median=3 (n=42,020)	mean=5.90, median=6 (n=501,247)	mean=6.33, median=6 (n=2,556,887)
`R2AR`	mean=3.45, median=3 (n=64,880)	mean=6.01, median=6 (n=600,312)	mean=6.34, median=6 (n=2,434,962)
`R2B`	mean=3.27, median=3 (n=62,041)	mean=6.05, median=6 (n=607,149)	mean=6.33, median=6 (n=2,430,964)
`R3AF`	mean=3.09, median=2 (n=50,534)	mean=5.91, median=6 (n=576,958)	mean=6.35, median=6 (n=2,472,662)
`R3AV`	mean=3.29, median=3 (n=62,755)	mean=5.93, median=6 (n=567,196)	mean=6.35, median=6 (n=2,470,203)
`R3B`	mean=2.93, median=2 (n=32,788)	mean=5.84, median=6 (n=619,807)	mean=6.35, median=6 (n=2,447,559)

Read counts per receptor

Receptor	n with count	mean	median	p90	p99	max
`R1`	751,569	57.30	25	118	348	513,819
`FcRn6`	716,069	68.27	31	137	397	318,477
`R2AH`	543,267	53.35	24	100	322	737,891
`R2AR`	665,192	43.85	19	78	231	899,653
`R2B`	669,190	58.31	26	102	301	1,098,765
`R3AF`	627,498	50.91	22	88	264	1,140,536
`R3AV`	629,957	69.44	32	130	371	940,365
`R3B`	652,606	50.16	26	109	289	54,390

Sequence diversity (per-position Shannon entropy)

Overall (sampled): n=200,000, mean entropy = 0.19 bits, max position = 2.84, positions above 1 bit = 13/232, zero-entropy positions = 91.

Sequences with all 8 labels: n=21,556, mean entropy = 0.11 bits, max position = 2.54, positions above 1 bit = 8/232, zero-entropy positions = 95.

Per receptor (binders vs non-binders)

Receptor	binders	non-binders
`R1`	n=100,000, mean H = 0.12	n=100,000, mean H = 0.17
`FcRn6`	n=100,000, mean H = 0.13	n=100,000, mean H = 0.17
`R2AH`	n=42,020, mean H = 0.11	n=100,000, mean H = 0.18
`R2AR`	n=64,880, mean H = 0.12	n=100,000, mean H = 0.18
`R2B`	n=62,041, mean H = 0.12	n=100,000, mean H = 0.18
`R3AF`	n=50,534, mean H = 0.11	n=100,000, mean H = 0.17
`R3AV`	n=62,755, mean H = 0.12	n=100,000, mean H = 0.18
`R3B`	n=32,788, mean H = 0.10	n=100,000, mean H = 0.17

Generated by scripts/data/analyze_dataset.py. Machine-readable companion: analysis.yaml in the same run directory.

Dataset analysis: paper