Cross-source overlap and quality-filter audit

Run id 20260528-083414_data_overlap_seed0 · 2026-05-28T08:34:14+00:00 · championsnet

What this report shows

Pairwise sequence overlap and per-receptor label agreement across paper, paper_refiltered, dms, first_experimental_validation.
The same analysis repeated under per-receptor bind_count + non_count thresholds, so the rows-lost-vs-noise-reduced trade-off on the filtered sources is visible.
Filters operate per receptor per row: a row keeps its R1 label if R1 passes the filter, even when its R3B label gets masked out.
Filters are applied only to paper_refiltered and dms. paper's count_R* is unreliable (no PCR-bias correction, per library author) and is kept unfiltered as a comparison anchor. first_experimental_validation has no read counts.

Sources

Rows with at least one label per source, by filter

Per-pair filter comparison

Filter	paper	paper_refiltered	dms	costain	first_experimental_validation
Baseline (no extra filter on paper_refiltered/dms/costain)	3,100,154	1,108,101	322,546	69,201	1,761
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	3,100,154	1,107,899	319,016	69,193	1,761
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	3,100,154	690,506	172,896	53,021	1,761

Each pair gets two tables: agreement % per receptor across filters, and the underlying comparable-row counts. Color coding: ≥ 95 %, 80 – 95 %, < 80 %.

`paper` × `paper_refiltered`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	99.80%	99.82%	99.53%	99.58%	99.70%	99.69%	99.72%	100.00%	99.73%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	99.81%	99.84%	99.53%	99.58%	99.70%	99.72%	99.74%	100.00%	99.74%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	99.74%	99.78%	99.27%	99.37%	99.47%	99.59%	99.64%	100.00%	99.61%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	total
Baseline (no extra filter on paper_refiltered/dms/costain)	287,585	281,705	144,887	201,783	149,438	225,579	229,171	266,933	1,787,081
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	240,913	235,174	109,335	156,451	107,517	177,468	186,195	222,796	1,435,849
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	151,375	149,630	58,915	86,557	52,602	102,748	113,530	142,044	857,401

`paper` × `dms`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	80.27%	86.78%	76.92%	77.92%	83.66%	73.19%	73.60%	71.09%	77.93%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	82.73%	83.89%	79.46%	79.63%	85.48%	75.26%	75.94%	72.23%	79.33%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	85.00%	82.37%	80.17%	81.56%	86.07%	75.68%	77.93%	72.14%	80.12%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	total
Baseline (no extra filter on paper_refiltered/dms/costain)	3,883	3,781	3,341	4,117	5,503	3,006	3,133	5,250	32,014
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	2,606	2,563	1,996	2,523	3,932	1,908	1,933	3,796	21,257
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	1,993	1,894	1,417	1,768	2,735	1,439	1,495	2,462	15,203

`paper` × `costain`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	94.58%	—	—	86.24%	86.46%	91.51%	—	—	89.70%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	94.55%	—	—	85.63%	86.32%	91.70%	—	—	89.55%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	95.16%	—	—	86.30%	87.08%	91.66%	—	—	90.05%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	R2AR	R2B	R3AF	total
Baseline (no extra filter on paper_refiltered/dms/costain)	12,721	7,810	7,532	10,856	38,919
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	12,302	6,603	6,398	9,846	35,149
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	10,667	5,564	5,401	8,262	29,894

`paper` × `first_experimental_validation`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	91.67%	80.00%	76.47%	72.22%	73.91%	72.73%	71.43%	100.00%	79.80%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	91.67%	80.00%	76.47%	72.22%	73.91%	72.73%	71.43%	100.00%	79.80%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	91.67%	80.00%	76.47%	72.22%	73.91%	72.73%	71.43%	100.00%	79.80%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	total
Baseline (no extra filter on paper_refiltered/dms/costain)	24	30	17	18	23	22	21	21	176
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	24	30	17	18	23	22	21	21	176
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	24	30	17	18	23	22	21	21	176

`paper_refiltered` × `dms`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	88.33%	93.53%	84.41%	84.93%	89.50%	81.52%	82.70%	82.86%	85.97%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	92.22%	96.86%	87.86%	88.89%	91.92%	84.25%	88.45%	83.67%	89.27%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	93.90%	97.95%	89.33%	92.06%	92.59%	85.52%	90.34%	85.71%	90.92%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	total
Baseline (no extra filter on paper_refiltered/dms/costain)	1,919	1,746	1,450	1,745	2,191	1,288	1,358	2,264	13,961
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	1,234	1,147	898	1,044	1,300	838	883	1,335	8,679
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	934	926	731	844	918	718	745	819	6,635

`paper_refiltered` × `costain`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	93.90%	—	—	83.65%	83.58%	93.51%	—	—	88.66%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	94.97%	—	—	82.36%	83.91%	93.25%	—	—	88.62%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	95.56%	—	—	81.98%	82.73%	92.34%	—	—	88.15%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	R2AR	R2B	R3AF	total
Baseline (no extra filter on paper_refiltered/dms/costain)	6,815	3,602	2,893	5,176	18,486
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	5,127	2,228	1,659	3,391	12,405
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	3,264	1,365	967	1,958	7,554

`paper_refiltered` × `first_experimental_validation`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	86.67%	83.33%	62.50%	77.78%	72.73%	90.91%	90.00%	100.00%	82.99%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	100.00%	85.71%	71.43%	75.00%	75.00%	90.91%	100.00%	100.00%	87.26%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	100.00%	100.00%	75.00%	80.00%	75.00%	80.00%	100.00%	100.00%	88.75%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	total
Baseline (no extra filter on paper_refiltered/dms/costain)	15	12	8	9	11	11	10	10	86
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	8	7	7	8	8	11	7	7	63
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	6	4	4	5	8	5	5	7	44

`dms` × `costain`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	86.20%	—	—	78.04%	82.95%	53.15%	—	—	75.08%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	87.65%	—	—	77.50%	81.50%	47.97%	—	—	73.66%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	89.26%	—	—	71.76%	82.63%	47.43%	—	—	72.77%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	R2AR	R2B	R3AF	total
Baseline (no extra filter on paper_refiltered/dms/costain)	761	583	868	461	2,673
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	583	280	427	246	1,536
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	419	170	259	175	1,023

`dms` × `first_experimental_validation`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%	100.00%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	total
Baseline (no extra filter on paper_refiltered/dms/costain)	2	2	1	2	1	1	1	2	12
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	2	1	1	1	1	1	1	2	10
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	2	1	1	1	1	1	1	2	10

`costain` × `first_experimental_validation`

Agreement % per receptor, by filter

Filter	R1	FcRn6	R2AH	R2AR	R2B	R3AF	R3AV	R3B	mean
Baseline (no extra filter on paper_refiltered/dms/costain)	100.00%	—	—	77.78%	77.78%	88.89%	—	—	86.11%
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	100.00%	—	—	77.78%	77.78%	88.89%	—	—	86.11%
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	100.00%	—	—	71.43%	66.67%	87.50%	—	—	81.40%

Comparable rows per receptor (both sources have non-missing label)

Filter	R1	R2AR	R2B	R3AF	total
Baseline (no extra filter on paper_refiltered/dms/costain)	9	9	9	9	36
bind+non ≥ 2 per receptor (paper_refiltered, dms, costain)	9	9	9	9	36
bind+non ≥ 3 per receptor (paper_refiltered, dms, costain)	6	7	6	8	27

`paper`	/cluster/project/reddy/sounder/data/FcRprofile_filtered/FcRprofile_filtered_data_full.csv
`paper_refiltered`	/cluster/project/reddy/sounder/combi_umi/combi_umi_dat.csv
`dms`	/cluster/project/reddy/sounder/dms/dms_dat.csv
`costain`	/cluster/project/reddy/sounder/costain/costain_dat.csv
`first_experimental_validation`	/cluster/project/reddy/sounder/experimental_validation/fcgpt_val_df_parsed.csv

Cross-source overlap and quality-filter audit

Sources

Rows with at least one label per source, by filter

Per-pair filter comparison

paper × paper_refiltered

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

paper × dms

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

paper × costain

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

paper × first_experimental_validation

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

paper_refiltered × dms

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

paper_refiltered × costain

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

paper_refiltered × first_experimental_validation

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

dms × costain

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

dms × first_experimental_validation

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

costain × first_experimental_validation

Agreement % per receptor, by filter

Comparable rows per receptor (both sources have non-missing label)

`paper` × `paper_refiltered`

`paper` × `dms`

`paper` × `costain`

`paper` × `first_experimental_validation`

`paper_refiltered` × `dms`

`paper_refiltered` × `costain`

`paper_refiltered` × `first_experimental_validation`

`dms` × `costain`

`dms` × `first_experimental_validation`

`costain` × `first_experimental_validation`