Experiment / E2C36R2YLEpisomal Plasmid MPRA

Round 4b margin-sampled MPRA

Active learning of enhancer and silencer regulatory grammar in photoreceptors

Two episomal MPRA sublibraries tested margin-uncertainty-selected sequences enriched for the silencer-versus-enhancer decision boundary, together with high-probability controls and repeats from Round 4a. Constructs were assayed in P0 CD-1 mouse retinal explants using the Rho basal promoter reporter.

Download processed table (CSV)Raw source data API documentation

Processed tables are specific to each experiment. Column names, units, measurements, and table structure are not standardized across the database. Check this experiment’s column definitions and quality-control notes before comparing or combining data.

Organism: Mouse
Taxonomy ID: NCBITaxon:10090
Biosample: UBERON:0000966
Reference genome: mm10
Design focus: Synthetic / Motif-focused
Region of interest: Not reported / not applicable

Perturbation & assay details

Basal / Untreated

Oligo-synthesized 164-bp test sequences with unique 9-bp cis-regulatory barcodes were cloned upstream of the rod-specific Rhodopsin (Rho) basal promoter driving DsRed in episomal pJK03-derived reporter plasmids. P0 CD-1 mouse retinas were electroporated, five retinas were pooled per biological replicate, and tissue was cultured for eight days; barcode RNA/cDNA counts were normalized to input plasmid DNA and basal or scrambled pseudobasal controls as specified by the authors.

Processed data

50 rows per page. Click a cell to inspect its full value.

Visible columns (46 of 46)

Row
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

Page 1 · 50 rows · More results available

Filters apply to this table only. The CSV download contains the complete processed table; filtered rows are available through the API.

Column dictionary · 46 definitions

element_id: Unique row identifier copied from the authors’ processed activity table.
label: Original sequence/construct label, including genomic coordinates and perturbation suffixes when present.
library: MPRA library name assigned by the authors (library1 through library9b).
data_batch_name: Authors’ active-learning batch annotation, such as Genomic, Round2, Round3a, Round4a, or Round4b.
sequence: 164-bp DNA test sequence used in the reporter construct.
sequence_length: Length of sequence in nucleotides; package QC required 164.
gc_fraction: Fraction of sequence bases that are G or C.
activity_score: Author basal-normalized mean MPRA activity on a linear scale; expression = 1 corresponds to basal activity.
activity_score_std: Author basal-normalized standard deviation across RNA replicate activity values.
activity_replicates: Number of RNA replicate activity values contributing to the author summary.
activity_log2: Log2 of the author basal-normalized MPRA activity score.
p_value_vs_basal: Author Welch-test p-value comparing the element with basal activity.
q_value_vs_basal: Author FDR-adjusted q-value for the comparison with basal activity.
lognormal_mu: Author fitted log-normal location parameter for replicate activity values.
lognormal_sigma: Author fitted log-normal scale parameter for replicate activity values.
activity_class: Author-discretized activity class: Silencer, Inactive, WeakEnhancer, or StrongEnhancer.
dna_count_sum: Sum of raw input-plasmid barcode counts assigned to this element across all matching barcodes.
rna_count_sum_rep1: Sum of raw RNA barcode counts for the first RNA sample in the library.
rna_count_sum_rep2: Sum of raw RNA barcode counts for the second RNA sample in the library.
rna_count_sum_rep3: Sum of raw RNA barcode counts for the third RNA sample in the library.
rna_count_sum_rep4: Sum of raw RNA barcode counts for the fourth RNA sample in the library, when available.
rna_count_sum_rep5: Sum of raw RNA barcode counts for the fifth RNA sample in the library, when available.
rna_count_sum_mean: Arithmetic mean of available raw RNA count sums across replicates.
rna_count_sum_cv: Population coefficient of variation of available raw RNA count sums; this is a package-level context metric, not the barcode-level author QC CV.
rna_nonzero_reps: Number of RNA samples with at least one matching raw count.
immediate_precursor: Sequence label from which the current perturbation was directly derived, when annotated by the authors.
original_seq: Original genomic sequence identifier associated with a perturbation, when annotated.
standard_seq: Boolean flag for an assay standard sequence.
scrambled: Boolean flag for a scrambled negative-control sequence.
rational_mutagenesis: Boolean flag for a hand-designed/rational motif perturbation.
ic_scan: Boolean flag for an information-content scan sequence.
original_genomic: Boolean flag for an original genomic sequence in libraries 1–2.
mut_all_crx: Boolean flag for a sequence with all CRX motifs mutated.
mut_shape: Boolean flag for a sequence with a CRX shape motif mutation.
entropy_sampling: Boolean flag for selection by entropy uncertainty sampling.
margin_sampling: Boolean flag for selection by margin uncertainty sampling.
random_sampling: Boolean flag for a random active-learning control.
high_conf_pilot: Boolean flag for a high-confidence pilot prediction sequence.
high_conf_cnn: Boolean flag for a high-confidence CNN prediction sequence.
l9_controls: Boolean flag for known-control sequences included in library 9.
l9_repeat_l8: Boolean flag for library 8 sequences repeated in library 9.
test_set: Boolean flag for the motif-mutagenesis test set.
derived_from_test_set_seq: Boolean flag for a sequence derived from a held-out test-set sequence.
cnn_validation_set: Boolean flag for membership in the CNN validation set.
svm_train: Boolean flag indicating inclusion in the authors’ SVM training set.
cnn_train: Boolean flag indicating inclusion in the authors’ CNN training set.

Quality control

Author QC required exact cBC/mBC context matches, at least 50 input-plasmid DNA counts per barcode, and RNA barcode coefficient of variation no greater than 0.8 across cDNA samples; counts were depth-normalized, RNA was normalized to plasmid DNA, barcodes were averaged to CRE activity, and activity was basal-normalized (or scrambled-pseudobasal-normalized when basal recovery was poor). The author-published summary retained only sequence measurements with more than one RNA replicate. Package QC additionally required unique element IDs, 164-bp A/C/G/T-only sequences, finite author activity statistics, at least one matching raw DNA count, and at least two nonzero raw RNA replicate totals. Retained 11945 of 12418 source rows; removed 473 rows: insufficient_raw_rna_counts=473.

Curation notes

Libraries9a–9b contain margin-sampled sequences, repeats of library8 elements, and known controls. These categories are retained together because they were measured in the same Round 4b library preparation and are explicitly annotated in the table.

Row
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

Row
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

Row
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50