Corpus Engine Towards a reproducible .science file.
← All .science files

pmid:34252925

scPNMF: sparse gene encoding of single cells to facilitate gene selection for targeted gene profiling.

Dongyuan Song · Kexin Li · Zachary Hemminger · Roy Wollman · Jingyi Jessica Li

Bioinformatics (Oxford, England) · 2021 doi:10.1093/bioinformatics/btab273

Paper

complete

Experiment

partial

Data

partial

Analysis

partial

Execution

absent

Execution evidence has not yet been added to this .science file. Explore its methods, data and open questions to plan the next step.
Record details sci_version · id · kind · revision · content_hash · schema · created_at · updated_at · generator
sci_version
1.2
id
pmid:34252925
kind
science-record
revision
7
content_hash
sha256:2291ce08acd660810d707a3ac69756d8a203323e8871592f8a9cc17061e19b13
schema
https://corpus.bntrg.co.uk/sci/1.0/sci.schema.json
created_at
2026-11-01T15:41:43Z
updated_at
2026-11-01T22:04:11Z
generator object · 2 fields
name
corpus-assay-engine
run_id
reading-sci-2026-11-01
paper object · 10 fields
state
complete
title
scPNMF: sparse gene encoding of single cells to facilitate gene selection for targeted gene profiling.
doi
10.1093/bioinformatics/btab273
pmid
34252925
authors array · 5 items
[0]
Dongyuan Song
[1]
Kexin Li
[2]
Zachary Hemminger
[3]
Roy Wollman
[4]
Jingyi Jessica Li
journal
Bioinformatics (Oxford, England)
year
2021
abstract
MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome information of individual cells. While scRNA-seq measures thousands of genes, researchers are often interested in only dozens to hundreds of genes for a closer study. Then, a question is how to select those informative genes from scRNA-seq data. Moreover, single-cell targeted gene profiling technologies are gaining popularity for their low costs, high sensitivity and extra (e.g. spatial) information; however, they typically can only measure up to a few hundred genes. Then another challenging question is how to select genes for targeted gene profiling based on existing scRNA-seq data. RESULTS: Here, we develop the single-cell Projective Non-negative Matrix Factorization (scPNMF) method to select informative genes from scRNA-seq data in an unsupervised way. Compared with existing gene selection methods, scPNMF has two advantages. First, its selected informative genes can better distinguish cell types. Second, it enables the alignment of new targeted gene profiling data with reference data in a low-dimensional space to facilitate the prediction of cell types in the new data. Technically, scPNMF modifies the PNMF algorithm for gene selection by changing the initialization and adding a basis selection step, which selects informative bases to distinguish cell types. We demonstrate that scPNMF outperforms the state-of-the-art gene selection methods on diverse scRNA-seq datasets. Moreover, we show that scPNMF can guide the design of targeted gene profiling experiments and the cell-type annotation on targeted gene profiling data. AVAILABILITY AND IMPLEMENTATION: The R package is open-access and available at https://github.com/JSB-UCLA/scPNMF. The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
claims array · 2 items
[0] object · 6 fields
id
claim-method
statement
Using the adjusted Rand index (ARI) as the metric of clustering accuracy, we calculate the ARI values of the three clustering methods on each dataset using 100 informative genes selected by each gene selection method, as 100 genes are commonly used in targeted gene profiling.
kind
methodological
target
null
evidence array · 1 item
[0] object · 3 fields
asset
source:pmid:34252925
locator
sections[15] "3.3 scPNMF outperforms state-of-the-art gene selection methods on diverse scRNA-seq datasets" paragraph 2
quote
Using the adjusted Rand index (ARI) as the metric of clustering accuracy, we calculate the ARI values of the three clustering methods on each dataset using 100 informative genes selected by each gene selection method, as 100 genes are commonly used in targeted gene profiling.
state
complete
[1] object · 6 fields
id
claim-data-access
statement
The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997.
kind
methodological
target
null
evidence array · 1 item
[0] object · 3 fields
asset
source:pmid:34252925
locator
sections[0] "Abstract" paragraph 3
quote
The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997.
state
complete
gaps array · 0 items
experiment object · 8 fields
state
partial
objective
null
steps array · 3 items
[0] object · 13 fields
id
read_published_estimate
order
1
name
Read the published estimate from the source
kind
other
state
complete
blocked_on
null
description
The claim, method and evidence span published with the source article.
duration_hours
null
parameters object · 0 fields
inputs array · 1 item
[0]
source:pmid:34252925
outputs array · 1 item
[0]
the published estimate and its evidence
instruments array · 0 items
reagents array · 0 items
[1] object · 13 fields
id
obtain_source_data
order
2
name
Obtain the source data from the named repository
kind
data-preparation
state
blocked
blocked_on
data:accession_absent
description
The GEO series accession resolves and its sample accessions are listed. This lane did not download the dataset or read its licence.
duration_hours
null
parameters object · 0 fields
inputs array · 0 items
outputs array · 1 item
[0]
the source data with its repository provenance
instruments array · 0 items
reagents array · 0 items
[2] object · 13 fields
id
run_analysis
order
3
name
Run the published analysis
kind
analysis
state
blocked
blocked_on
method:machine_learning_random_forest
description
Apply the paper's statistical method to the resolved series and compare the result with the published target.
duration_hours
null
parameters object · 0 fields
inputs array · 0 items
outputs array · 1 item
[0]
the replicated result and its comparison
instruments array · 0 items
reagents array · 0 items
materials array · 0 items
instruments array · 0 items
reagents array · 0 items
conditions object · 0 fields
gaps array · 2 items
[0]
step:run_analysis blocked on method:machine_learning_random_forest
[1]
step:obtain_source_data blocked on data:accession_absent
data object · 3 fields
state
partial
assets array · 3 items
[0] object · 15 fields
id
source:pmid:34252925
role
manuscript
format
xml
media_type
application/xml
uri
https://doi.org/10.1093/bioinformatics/btab273
sha256
sha256:43eb22792366393f8f71ba4bba5441f185e357e376184106c07401ea3f8bc15c
bytes
39767
access
open
state
resolved
reader
null
reader_status
reader_required
identifier
10.1093/bioinformatics/btab273
version
null
note
Full article fetched from PMC; the checksum covers the rendered section body, not the original XML bytes.
structure object · 4 fields
pmcid
PMC8275345
source
europe_pmc_fulltext
coverage
whole_body_without_references
text_hash_scope
rendered_section_body
[1] object · 16 fields
id
asset:table:pmid:34252925:1
role
table
format
jats-table
media_type
application/vnd.jats.table+json
uri
null
sha256
sha256:6b45ffc0544d6672a44c5faa49d8a963d18ff7f250f1218e9aca51def79ba240
bytes
505
access
open
state
verified
reader
read_table
reader_status
available
identifier
section[3]/table-wrap[1]#btab273-T1
version
null
note
Comparison of the properties of PNMF, PCA and NMF
structure object · 5 fields
pmcid
PMC8275345
section_index
3
section_title
2.1 scPNMF Step I: PNMF
row_count
4
column_count
6
table object · 8 fields
label
Table 1
caption
Comparison of the properties of PNMF, PCA and NMF
rows array · 4 items
[0] array · 6 items
[0]
[1]
Optimization problem
[2]
Non-negativity
[3]
Sparsity
[4]
Mutual exclusiveness
[5]
New data projection
[1] array · 6 items
[0]
PNMF
[1]
minW||X−WWTX|| s.t. W≥0
[2]
Yes
[3]
Very high
[4]
Very high
[5]
Yes
[2] array · 6 items
[0]
PCA
[1]
minW||X−WWTX|| s.t. WTW=I
[2]
No
[3]
Low
[4]
Low
[5]
Yes
[3] array · 6 items
[0]
NMF
[1]
minW,H||X−WH|| s.t. W,H≥0
[2]
Yes
[3]
High
[4]
High
[5]
No
row_count
4
column_count
6
numeric_cells
2
header_rows
1
locator
section[3]/table-wrap[1]#btab273-T1
[2] object · 16 fields
id
asset:table:pmid:34252925:2
role
table
format
jats-table
media_type
application/vnd.jats.table+json
uri
null
sha256
sha256:c5d44d2f51e07f68f3cf45de8fac953625ec34d70da5c6244f67c36d6e527dea
bytes
2061
access
open
state
verified
reader
read_table
reader_status
available
identifier
section[16]/table-wrap[1]#btab273-T2
version
null
note
Prediction accuracy of cell types based on 100 informative genes selected by 12 gene selection methods in the two case studies with paired reference scRNA-seq data and targeted gene profiling data
structure object · 5 fields
pmcid
PMC8275345
section_index
16
section_title
3.4 scPNMF guides targeted gene profiling experimental design and cell-type prediction
row_count
14
column_count
8
table object · 8 fields
label
Table 2.
caption
Prediction accuracy of cell types based on 100 informative genes selected by 12 gene selection methods in the two case studies with paired reference scRNA-seq data and targeted gene profiling data
rows array · 14 items
[0] array · 4 items
[0]
Method
[1]
Zheng8
[2]
PBMC
[3]
Average Accuracy
[1] array · 8 items
[0]
[1]
RandomForest
[2]
KNN
[3]
SVM
[4]
RandomForest
[5]
KNN
[6]
SVM
[7]
[2] array · 8 items
[0]
scPNMF
[1]
0.85 (0.83,0.87)
[2]
0.80 (0.78,0.83)
[3]
0.87 (0.85,0.89)
[4]
0.84 (0.79,0.88)
[5]
0.84 (0.79,0.88)
[6]
0.67 (0.61,0.73)
[7]
0.81
[3] array · 8 items
[0]
M3Drop
[1]
0.85 (0.83,0.87)
[2]
0.80 (0.77,0.83)
[3]
0.87 (0.84,0.89)
[4]
0.84 (0.79,0.88)
[5]
0.77 (0.71,0.82)
[6]
0.63 (0.57,0.69)
[7]
0.79
[4] array · 8 items
[0]
SeuratDISP
[1]
0.84 (0.81,0.86)
[2]
0.78 (0.75,0.81)
[3]
0.86 (0.84,0.88)
[4]
0.80 (0.75,0.84)
[5]
0.75 (0.70,0.80)
[6]
0.64 (0.58,0.70)
[7]
0.78
[5] array · 8 items
[0]
corFS
[1]
0.80 (0.77,0.82)
[2]
0.75 (0.73,0.78)
[3]
0.82 (0.80,0.85)
[4]
0.82 (0.77,0.86)
[5]
0.81 (0.76,0.86)
[6]
0.62 (0.56,0.68)
[7]
0.77
[6] array · 8 items
[0]
GiniClust
[1]
0.86 (0.83,0.88)
[2]
0.79 (0.76,0.81)
[3]
0.86 (0.83,0.88)
[4]
0.80 (0.75,0.84)
[5]
0.76 (0.71,0.81)
[6]
0.53 (0.47,0.60)
[7]
0.75
[7] array · 8 items
[0]
scran
[1]
0.79 (0.76,0.81)
[2]
0.72 (0.69,0.75)
[3]
0.82 (0.80,0.85)
[4]
0.78 (0.72,0.82)
[5]
0.73 (0.67,0.78)
[6]
0.67 (0.61,0.72)
[7]
0.75
[8] array · 8 items
[0]
SeuratMVP
[1]
0.83 (0.81,0.85)
[2]
0.77 (0.74,0.80)
[3]
0.85 (0.82,0.87)
[4]
0.82 (0.77,0.86)
[5]
0.74 (0.69,0.79)
[6]
0.47 (0.40,0.53)
[7]
0.74
[9] array · 8 items
[0]
Scanpy
[1]
0.79 (0.77,0.82)
[2]
0.71 (0.68,0.74)
[3]
0.80 (0.78,0.83)
[4]
0.80 (0.75,0.84)
[5]
0.76 (0.71,0.81)
[6]
0.52 (0.46,0.58)
[7]
0.73
[10] array · 8 items
[0]
SCMarker
[1]
0.77 (0.74,0.79)
[2]
0.68 (0.65,0.71)
[3]
0.74 (0.71,0.77)
[4]
0.77 (0.71,0.81)
[5]
0.71 (0.65,0.76)
[6]
0.45 (0.39,0.52)
[7]
0.69
[11] array · 8 items
[0]
SeuratVST
[1]
0.73 (0.70,0.76)
[2]
0.68 (0.65,0.71)
[3]
0.75 (0.73,0.78)
[4]
0.74 (0.68,0.79)
[5]
0.68 (0.63,0.74)
[6]
0.40 (0.34,0.46)
[7]
0.67
[12] array · 8 items
[0]
DANB
[1]
0.71 (0.68,0.73)
[2]
0.69 (0.66,0.71)
[3]
0.75 (0.73,0.78)
[4]
0.73 (0.67,0.78)
[5]
0.74 (0.68,0.79)
[6]
0.28 (0.23,0.34)
[7]
0.65
[13] array · 8 items
[0]
irlbaPcaFS
[1]
0.68 (0.65,0.71)
[2]
0.61 (0.58,0.64)
[3]
0.71 (0.68,0.74)
[4]
0.71 (0.65,0.76)
[5]
0.77 (0.71,0.82)
[6]
0.16 (0.12,0.21)
[7]
0.61
row_count
14
column_count
8
numeric_cells
84
header_rows
2
locator
section[16]/table-wrap[1]#btab273-T2
gaps array · 0 items
analysis object · 4 fields
state
partial
methods array · 5 items
[0] object · 10 fields
id
method-1
family
machine_learning
specific
random_forest
label
Machine learning
dsl_operations array · 0 items
missing_operations array · 1 item
[0]
machine_learning_random_forest
expressible
false
parameters object · 0 fields
data_shapes array · 1 item
[0]
labelled_feature_matrix
evidence object · 3 fields
score
0.8
band
high
quote
random forest
[1] object · 10 fields
id
method-2
family
machine_learning
specific
support_vector_machine
label
Machine learning
dsl_operations array · 3 items
[0]
train_classifier
[1]
cross_validate
[2]
roc_auc
missing_operations array · 0 items
expressible
true
parameters object · 0 fields
data_shapes array · 1 item
[0]
labelled_feature_matrix
evidence object · 3 fields
score
0.8
band
high
quote
support vector machine
[2] object · 10 fields
id
method-3
family
clustering
specific
kmeans
label
Clustering
dsl_operations array · 1 item
[0]
kmeans_fit
missing_operations array · 0 items
expressible
true
parameters object · 0 fields
data_shapes array · 1 item
[0]
feature_matrix
evidence object · 3 fields
score
0.8
band
high
quote
K-means
[3] object · 10 fields
id
method-4
family
clustering
specific
hierarchical
label
Clustering
dsl_operations array · 1 item
[0]
hierarchical_cluster
missing_operations array · 0 items
expressible
true
parameters object · 0 fields
data_shapes array · 1 item
[0]
feature_matrix
evidence object · 3 fields
score
0.8
band
high
quote
hierarchical clustering
[4] object · 10 fields
id
method-5
family
null
specific
null
label
adjusted Rand index (ARI)
dsl_operations array · 0 items
missing_operations array · 0 items
expressible
null
parameters object · 0 fields
data_shapes array · 0 items
evidence object · 3 fields
score
1
band
verbatim
quote
Using the adjusted Rand index (ARI) as the metric of clustering accuracy, we calculate the ARI values of the three clustering methods on each dataset using 100 informative genes selected by each gene selection method, as 100 genes are commonly used in targeted gene profiling.
targets array · 0 items
gaps array · 2 items
[0]
method.reader_named_unmapped
[1]
method.machine_learning_random_forest
capability object · 3 fields
state
complete
requirements array · 1 item
[0] object · 6 fields
capability_id
capability:computational_analysis
label
Computational analysis platform
assay_families array · 9 items
[0]
meta_analysis
[1]
machine_learning
[2]
survival_analysis
[3]
statistical_regression
[4]
pca
[5]
molecular_docking
[6]
network_analysis
[7]
differential_expression
[8]
gwas
availability_state
owned
tier_unlocked
R1
candidate
Automated Assays Laravel application and simulated adapters
gaps array · 0 items
readiness object · 8 fields
state
blocked
tier
R4
blocking_axis
data
blocked_on
data:accession_absent
axes object · 3 fields
method object · 5 fields
tier
R1
expressible
false
required_operations array · 0 items
missing_operations array · 1 item
[0]
machine_learning_random_forest
basis
The paper names the method it used and the corpus's method extractor maps it onto the DSL catalogue: random_forest.
capability object · 4 fields
tier
R1
capability_id
capability:computational_analysis
availability_state
owned
basis
The computational analysis is owned. The physical bench the paper used is not, so a physical validation cannot be run here.
data object · 6 fields
tier
R4
reachable
false
accessions array · 0 items
resolved_accessions array · 0 items
statements array · 1 item
[0]
The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997.
basis
The paper names no accession to its source data, so the data has to come from the authors.
can_run object · 3 fields
dsl_operations array · 0 items
assets array · 3 items
[0]
source:pmid:34252925
[1]
asset:table:pmid:34252925:1
[2]
asset:table:pmid:34252925:2
steps array · 0 items
must_build object · 5 fields
dsl_operations array · 1 item
[0] object · 3 fields
operation
machine_learning_random_forest
unlocks_papers
null
priority
null
converters array · 0 items
readers array · 0 items
capabilities array · 0 items
data_gaps array · 1 item
[0]
data:geo:None
gaps array · 1 item
[0]
data:accession_absent
review object · 5 fields
state
partial
machine_critique object · 5 fields
kind
extraction_review
reader object · 5 fields
model
MiniMax-M3.1-Flash
provider
minimax
provider_family
minimax
read_at
2026-10-31T18:32:06Z
independence
single_reader
evidence_check object · 2 fields
method
verbatim
availability
verbatim
calls
3
usage object · 3 fields
prompt_tokens
29081
completion_tokens
961
cached_tokens
792
human_critique
null
stages array · 1 item
[0] object · 4 fields
id
extraction
state
complete
reader
MiniMax-M3.1-Flash
provider
minimax
gaps array · 3 items
[0]
review.independent_reader
[1]
review.machine_critique
[2]
review.human_critique
execution object · 5 fields
state
absent
status
not_run
mode
null
runs array · 0 items
gaps array · 1 item
[0]
execution.not_run
provenance object · 7 fields
entities array · 3 items
[0] object · 3 fields
id
pmid:34252925
type
paper
label
pmid:34252925
[1] object · 3 fields
id
source:pmid:34252925
type
manuscript
label
PMID 34252925 full text
[2] object · 3 fields
id
data:geo:None
type
dataset
label
GEO None
activities array · 3 items
[0] object · 4 fields
id
activity-reading-sci-2026-11-01
type
classification
used array · 3 items
[0] object · 3 fields
id
pmid:34252925
type
paper
label
pmid:34252925
[1] object · 3 fields
id
source:pmid:34252925
type
manuscript
label
PMID 34252925 full text
[2] object · 3 fields
id
data:geo:None
type
dataset
label
GEO None
agent
agent:corpus-assay-engine
[1] object · 4 fields
id
activity-read-34252925
type
extraction
used array · 1 item
[0]
source:pmid:34252925
agent
agent:minimax-reader
[2] object · 4 fields
id
activity-resolve-None
type
accession_resolution
used array · 1 item
[0]
data:geo:None
agent
agent:corpus-assay-engine
agents array · 2 items
[0] object · 3 fields
id
agent:corpus-assay-engine
type
software
label
Corpus assay engine
[1] object · 3 fields
id
agent:minimax-reader
type
software
label
MiniMax literature reader
source_run_id
reading-sci-2026-11-01
dsl_catalogue_fingerprint
343fb30756ba6ec3
capability_table_fingerprint
2cb3e42f047d2ecf
snapshot_id
ba34b49490d9bfaf
lineage object · 4 fields
parent_id
pmid:34252925
parent_hash
sha256:162437831476e862a5430a5743faa4d8ed1648511f7e230faf5620cbb39fad8f
fork_of
null
generation
6
gaps array · 8 items
[0]
execution.not_run
[1]
method.machine_learning_random_forest
[2]
method.reader_named_unmapped
[3]
review.human_critique
[4]
review.independent_reader
[5]
review.machine_critique
[6]
step:obtain_source_data blocked on data:accession_absent
[7]
step:run_analysis blocked on method:machine_learning_random_forest
assays array · 3 items
[0] object · 11 fields
role
data_generation
definition_id
rna_seq
definition_version
99e1195c7d6f
definition_digest
null
required_operations array · 0 items
support_state
partial
source_hashes object · 1 field
assay_lexicon
99e1195c7d6f
confidence
0.99
evidence array · 2 items
[0]
MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome inf
[1]
ll RNA sequencing (scRNA-seq) captures whole transcriptome information of individual cells. While scRNA
order_basis
paper_text_order:103
id
assay:1
[1] object · 11 fields
role
data_generation
definition_id
scrna_seq
definition_version
99e1195c7d6f
definition_digest
null
required_operations array · 0 items
support_state
unsupported
source_hashes object · 1 field
assay_lexicon
99e1195c7d6f
confidence
0.99
evidence array · 4 items
[0]
MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome inf
[1]
MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome information of
[2]
ptome information of individual cells. While scRNA-seq measures thousands of genes, researchers are
[3]
s how to select those informative genes from scRNA-seq data. Moreover, single-cell targeted gene pr
order_basis
paper_text_order:103
id
assay:2
[2] object · 11 fields
role
analysis
definition_id
machine_learning
definition_version
f2132e96ae6c7a4e
definition_digest
null
required_operations array · 0 items
support_state
partial
source_hashes object · 1 field
dsl_catalogue
f2132e96ae6c7a4e
confidence
0.8
evidence array · 1 item
[0]
random forest
order_basis
consumes_earlier_output
id
assay:3
Complete file 
{"sci_version": "1.2", "schema": "https://corpus.bntrg.co.uk/sci/1.0/sci.schema.json", "id": "pmid:34252925", "kind": "science-record", "revision": 7, "content_hash": "sha256:2291ce08acd660810d707a3ac69756d8a203323e8871592f8a9cc17061e19b13", "created_at": "2026-11-01T15:41:43Z", "updated_at": "2026-11-01T22:04:11Z", "generator": {"name": "corpus-assay-engine", "run_id": "reading-sci-2026-11-01"}, "paper": {"state": "complete", "title": "scPNMF: sparse gene encoding of single cells to facilitate gene selection for targeted gene profiling.", "doi": "10.1093/bioinformatics/btab273", "pmid": "34252925", "authors": ["Dongyuan Song", "Kexin Li", "Zachary Hemminger", "Roy Wollman", "Jingyi Jessica Li"], "journal": "Bioinformatics (Oxford, England)", "year": 2021, "abstract": "MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome information of individual cells. While scRNA-seq measures thousands of genes, researchers are often interested in only dozens to hundreds of genes for a closer study. Then, a question is how to select those informative genes from scRNA-seq data. Moreover, single-cell targeted gene profiling technologies are gaining popularity for their low costs, high sensitivity and extra (e.g. spatial) information; however, they typically can only measure up to a few hundred genes. Then another challenging question is how to select genes for targeted gene profiling based on existing scRNA-seq data.\nRESULTS: Here, we develop the single-cell Projective Non-negative Matrix Factorization (scPNMF) method to select informative genes from scRNA-seq data in an unsupervised way. Compared with existing gene selection methods, scPNMF has two advantages. First, its selected informative genes can better distinguish cell types. Second, it enables the alignment of new targeted gene profiling data with reference data in a low-dimensional space to facilitate the prediction of cell types in the new data. Technically, scPNMF modifies the PNMF algorithm for gene selection by changing the initialization and adding a basis selection step, which selects informative bases to distinguish cell types. We demonstrate that scPNMF outperforms the state-of-the-art gene selection methods on diverse scRNA-seq datasets. Moreover, we show that scPNMF can guide the design of targeted gene profiling experiments and the cell-type annotation on targeted gene profiling data.\nAVAILABILITY AND IMPLEMENTATION: The R package is open-access and available at https://github.com/JSB-UCLA/scPNMF. The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997.\nSUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.", "claims": [{"id": "claim-method", "statement": "Using the adjusted Rand index (ARI) as the metric of clustering accuracy, we calculate the ARI values of the three clustering methods on each dataset using 100 informative genes selected by each gene selection method, as 100 genes are commonly used in targeted gene profiling.", "kind": "methodological", "target": null, "evidence": [{"asset": "source:pmid:34252925", "locator": "sections[15] \"3.3 scPNMF outperforms state-of-the-art gene selection methods on diverse scRNA-seq datasets\" paragraph 2", "quote": "Using the adjusted Rand index (ARI) as the metric of clustering accuracy, we calculate the ARI values of the three clustering methods on each dataset using 100 informative genes selected by each gene selection method, as 100 genes are commonly used in targeted gene profiling."}], "state": "complete"}, {"id": "claim-data-access", "statement": "The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997.", "kind": "methodological", "target": null, "evidence": [{"asset": "source:pmid:34252925", "locator": "sections[0] \"Abstract\" paragraph 3", "quote": "The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997."}], "state": "complete"}], "gaps": []}, "experiment": {"state": "partial", "objective": null, "steps": [{"id": "read_published_estimate", "order": 1, "name": "Read the published estimate from the source", "kind": "other", "state": "complete", "blocked_on": null, "description": "The claim, method and evidence span published with the source article.", "duration_hours": null, "parameters": {}, "inputs": ["source:pmid:34252925"], "outputs": ["the published estimate and its evidence"], "instruments": [], "reagents": []}, {"id": "obtain_source_data", "order": 2, "name": "Obtain the source data from the named repository", "kind": "data-preparation", "state": "blocked", "blocked_on": "data:accession_absent", "description": "The GEO series accession resolves and its sample accessions are listed. This lane did not download the dataset or read its licence.", "duration_hours": null, "parameters": {}, "inputs": [], "outputs": ["the source data with its repository provenance"], "instruments": [], "reagents": []}, {"id": "run_analysis", "order": 3, "name": "Run the published analysis", "kind": "analysis", "state": "blocked", "blocked_on": "method:machine_learning_random_forest", "description": "Apply the paper's statistical method to the resolved series and compare the result with the published target.", "duration_hours": null, "parameters": {}, "inputs": [], "outputs": ["the replicated result and its comparison"], "instruments": [], "reagents": []}], "materials": [], "instruments": [], "reagents": [], "conditions": {}, "gaps": ["step:run_analysis blocked on method:machine_learning_random_forest", "step:obtain_source_data blocked on data:accession_absent"]}, "data": {"state": "partial", "assets": [{"id": "source:pmid:34252925", "role": "manuscript", "format": "xml", "media_type": "application/xml", "uri": "https://doi.org/10.1093/bioinformatics/btab273", "sha256": "sha256:43eb22792366393f8f71ba4bba5441f185e357e376184106c07401ea3f8bc15c", "bytes": 39767, "access": "open", "state": "resolved", "reader": null, "reader_status": "reader_required", "identifier": "10.1093/bioinformatics/btab273", "version": null, "note": "Full article fetched from PMC; the checksum covers the rendered section body, not the original XML bytes.", "structure": {"pmcid": "PMC8275345", "source": "europe_pmc_fulltext", "coverage": "whole_body_without_references", "text_hash_scope": "rendered_section_body"}}, {"id": "asset:table:pmid:34252925:1", "role": "table", "format": "jats-table", "media_type": "application/vnd.jats.table+json", "uri": null, "sha256": "sha256:6b45ffc0544d6672a44c5faa49d8a963d18ff7f250f1218e9aca51def79ba240", "bytes": 505, "access": "open", "state": "verified", "reader": "read_table", "reader_status": "available", "identifier": "section[3]/table-wrap[1]#btab273-T1", "version": null, "note": "Comparison of the properties of PNMF, PCA and NMF", "structure": {"pmcid": "PMC8275345", "section_index": 3, "section_title": "2.1 scPNMF Step I: PNMF", "row_count": 4, "column_count": 6}, "table": {"label": "Table 1", "caption": "Comparison of the properties of PNMF, PCA and NMF", "rows": [["", "Optimization problem", "Non-negativity", "Sparsity", "Mutual exclusiveness", "New data projection"], ["PNMF", "minW||X−WWTX|| s.t. W≥0", "Yes", "Very high", "Very high", "Yes"], ["PCA", "minW||X−WWTX|| s.t. WTW=I", "No", "Low", "Low", "Yes"], ["NMF", "minW,H||X−WH|| s.t. W,H≥0", "Yes", "High", "High", "No"]], "row_count": 4, "column_count": 6, "numeric_cells": 2, "header_rows": 1, "locator": "section[3]/table-wrap[1]#btab273-T1"}}, {"id": "asset:table:pmid:34252925:2", "role": "table", "format": "jats-table", "media_type": "application/vnd.jats.table+json", "uri": null, "sha256": "sha256:c5d44d2f51e07f68f3cf45de8fac953625ec34d70da5c6244f67c36d6e527dea", "bytes": 2061, "access": "open", "state": "verified", "reader": "read_table", "reader_status": "available", "identifier": "section[16]/table-wrap[1]#btab273-T2", "version": null, "note": "Prediction accuracy of cell types based on 100 informative genes selected by 12 gene selection methods in the two case studies with paired reference scRNA-seq data and targeted gene profiling data", "structure": {"pmcid": "PMC8275345", "section_index": 16, "section_title": "3.4 scPNMF guides targeted gene profiling experimental design and cell-type prediction", "row_count": 14, "column_count": 8}, "table": {"label": "Table 2.", "caption": "Prediction accuracy of cell types based on 100 informative genes selected by 12 gene selection methods in the two case studies with paired reference scRNA-seq data and targeted gene profiling data", "rows": [["Method", "Zheng8", "PBMC", "Average Accuracy"], ["", "RandomForest", "KNN", "SVM", "RandomForest", "KNN", "SVM", ""], ["scPNMF", "0.85 (0.83,0.87)", "0.80 (0.78,0.83)", "0.87 (0.85,0.89)", "0.84 (0.79,0.88)", "0.84 (0.79,0.88)", "0.67 (0.61,0.73)", "0.81"], ["M3Drop", "0.85 (0.83,0.87)", "0.80 (0.77,0.83)", "0.87 (0.84,0.89)", "0.84 (0.79,0.88)", "0.77 (0.71,0.82)", "0.63 (0.57,0.69)", "0.79"], ["SeuratDISP", "0.84 (0.81,0.86)", "0.78 (0.75,0.81)", "0.86 (0.84,0.88)", "0.80 (0.75,0.84)", "0.75 (0.70,0.80)", "0.64 (0.58,0.70)", "0.78"], ["corFS", "0.80 (0.77,0.82)", "0.75 (0.73,0.78)", "0.82 (0.80,0.85)", "0.82 (0.77,0.86)", "0.81 (0.76,0.86)", "0.62 (0.56,0.68)", "0.77"], ["GiniClust", "0.86 (0.83,0.88)", "0.79 (0.76,0.81)", "0.86 (0.83,0.88)", "0.80 (0.75,0.84)", "0.76 (0.71,0.81)", "0.53 (0.47,0.60)", "0.75"], ["scran", "0.79 (0.76,0.81)", "0.72 (0.69,0.75)", "0.82 (0.80,0.85)", "0.78 (0.72,0.82)", "0.73 (0.67,0.78)", "0.67 (0.61,0.72)", "0.75"], ["SeuratMVP", "0.83 (0.81,0.85)", "0.77 (0.74,0.80)", "0.85 (0.82,0.87)", "0.82 (0.77,0.86)", "0.74 (0.69,0.79)", "0.47 (0.40,0.53)", "0.74"], ["Scanpy", "0.79 (0.77,0.82)", "0.71 (0.68,0.74)", "0.80 (0.78,0.83)", "0.80 (0.75,0.84)", "0.76 (0.71,0.81)", "0.52 (0.46,0.58)", "0.73"], ["SCMarker", "0.77 (0.74,0.79)", "0.68 (0.65,0.71)", "0.74 (0.71,0.77)", "0.77 (0.71,0.81)", "0.71 (0.65,0.76)", "0.45 (0.39,0.52)", "0.69"], ["SeuratVST", "0.73 (0.70,0.76)", "0.68 (0.65,0.71)", "0.75 (0.73,0.78)", "0.74 (0.68,0.79)", "0.68 (0.63,0.74)", "0.40 (0.34,0.46)", "0.67"], ["DANB", "0.71 (0.68,0.73)", "0.69 (0.66,0.71)", "0.75 (0.73,0.78)", "0.73 (0.67,0.78)", "0.74 (0.68,0.79)", "0.28 (0.23,0.34)", "0.65"], ["irlbaPcaFS", "0.68 (0.65,0.71)", "0.61 (0.58,0.64)", "0.71 (0.68,0.74)", "0.71 (0.65,0.76)", "0.77 (0.71,0.82)", "0.16 (0.12,0.21)", "0.61"]], "row_count": 14, "column_count": 8, "numeric_cells": 84, "header_rows": 2, "locator": "section[16]/table-wrap[1]#btab273-T2"}}], "gaps": []}, "analysis": {"state": "partial", "methods": [{"id": "method-1", "family": "machine_learning", "specific": "random_forest", "label": "Machine learning", "dsl_operations": [], "missing_operations": ["machine_learning_random_forest"], "expressible": false, "parameters": {}, "data_shapes": ["labelled_feature_matrix"], "evidence": {"score": 0.8, "band": "high", "quote": "random forest"}}, {"id": "method-2", "family": "machine_learning", "specific": "support_vector_machine", "label": "Machine learning", "dsl_operations": ["train_classifier", "cross_validate", "roc_auc"], "missing_operations": [], "expressible": true, "parameters": {}, "data_shapes": ["labelled_feature_matrix"], "evidence": {"score": 0.8, "band": "high", "quote": "support vector machine"}}, {"id": "method-3", "family": "clustering", "specific": "kmeans", "label": "Clustering", "dsl_operations": ["kmeans_fit"], "missing_operations": [], "expressible": true, "parameters": {}, "data_shapes": ["feature_matrix"], "evidence": {"score": 0.8, "band": "high", "quote": "K-means"}}, {"id": "method-4", "family": "clustering", "specific": "hierarchical", "label": "Clustering", "dsl_operations": ["hierarchical_cluster"], "missing_operations": [], "expressible": true, "parameters": {}, "data_shapes": ["feature_matrix"], "evidence": {"score": 0.8, "band": "high", "quote": "hierarchical clustering"}}, {"id": "method-5", "family": null, "specific": null, "label": "adjusted Rand index (ARI)", "dsl_operations": [], "missing_operations": [], "expressible": null, "parameters": {}, "data_shapes": [], "evidence": {"score": 1.0, "band": "verbatim", "quote": "Using the adjusted Rand index (ARI) as the metric of clustering accuracy, we calculate the ARI values of the three clustering methods on each dataset using 100 informative genes selected by each gene selection method, as 100 genes are commonly used in targeted gene profiling."}}], "targets": [], "gaps": ["method.reader_named_unmapped", "method.machine_learning_random_forest"]}, "capability": {"state": "complete", "requirements": [{"capability_id": "capability:computational_analysis", "label": "Computational analysis platform", "assay_families": ["meta_analysis", "machine_learning", "survival_analysis", "statistical_regression", "pca", "molecular_docking", "network_analysis", "differential_expression", "gwas"], "availability_state": "owned", "tier_unlocked": "R1", "candidate": "Automated Assays Laravel application and simulated adapters"}], "gaps": []}, "readiness": {"state": "blocked", "tier": "R4", "blocking_axis": "data", "blocked_on": "data:accession_absent", "axes": {"method": {"tier": "R1", "expressible": false, "required_operations": [], "missing_operations": ["machine_learning_random_forest"], "basis": "The paper names the method it used and the corpus's method extractor maps it onto the DSL catalogue: random_forest."}, "capability": {"tier": "R1", "capability_id": "capability:computational_analysis", "availability_state": "owned", "basis": "The computational analysis is owned. The physical bench the paper used is not, so a physical validation cannot be run here."}, "data": {"tier": "R4", "reachable": false, "accessions": [], "resolved_accessions": [], "statements": ["The data used in this work are available at Zenodo: https://doi.org/10.5281/zenodo.4797997."], "basis": "The paper names no accession to its source data, so the data has to come from the authors."}}, "can_run": {"dsl_operations": [], "assets": ["source:pmid:34252925", "asset:table:pmid:34252925:1", "asset:table:pmid:34252925:2"], "steps": []}, "must_build": {"dsl_operations": [{"operation": "machine_learning_random_forest", "unlocks_papers": null, "priority": null}], "converters": [], "readers": [], "capabilities": [], "data_gaps": ["data:geo:None"]}, "gaps": ["data:accession_absent"]}, "review": {"state": "partial", "machine_critique": {"kind": "extraction_review", "reader": {"model": "MiniMax-M3.1-Flash", "provider": "minimax", "provider_family": "minimax", "read_at": "2026-10-31T18:32:06Z", "independence": "single_reader"}, "evidence_check": {"method": "verbatim", "availability": "verbatim"}, "calls": 3, "usage": {"prompt_tokens": 29081, "completion_tokens": 961, "cached_tokens": 792}}, "human_critique": null, "stages": [{"id": "extraction", "state": "complete", "reader": "MiniMax-M3.1-Flash", "provider": "minimax"}], "gaps": ["review.independent_reader", "review.machine_critique", "review.human_critique"]}, "execution": {"state": "absent", "status": "not_run", "mode": null, "runs": [], "gaps": ["execution.not_run"]}, "provenance": {"entities": [{"id": "pmid:34252925", "type": "paper", "label": "pmid:34252925"}, {"id": "source:pmid:34252925", "type": "manuscript", "label": "PMID 34252925 full text"}, {"id": "data:geo:None", "type": "dataset", "label": "GEO None"}], "activities": [{"id": "activity-reading-sci-2026-11-01", "type": "classification", "used": [{"id": "pmid:34252925", "type": "paper", "label": "pmid:34252925"}, {"id": "source:pmid:34252925", "type": "manuscript", "label": "PMID 34252925 full text"}, {"id": "data:geo:None", "type": "dataset", "label": "GEO None"}], "agent": "agent:corpus-assay-engine"}, {"id": "activity-read-34252925", "type": "extraction", "used": ["source:pmid:34252925"], "agent": "agent:minimax-reader"}, {"id": "activity-resolve-None", "type": "accession_resolution", "used": ["data:geo:None"], "agent": "agent:corpus-assay-engine"}], "agents": [{"id": "agent:corpus-assay-engine", "type": "software", "label": "Corpus assay engine"}, {"id": "agent:minimax-reader", "type": "software", "label": "MiniMax literature reader"}], "source_run_id": "reading-sci-2026-11-01", "dsl_catalogue_fingerprint": "343fb30756ba6ec3", "capability_table_fingerprint": "2cb3e42f047d2ecf", "snapshot_id": "ba34b49490d9bfaf"}, "lineage": {"parent_id": "pmid:34252925", "parent_hash": "sha256:162437831476e862a5430a5743faa4d8ed1648511f7e230faf5620cbb39fad8f", "fork_of": null, "generation": 6}, "gaps": ["execution.not_run", "method.machine_learning_random_forest", "method.reader_named_unmapped", "review.human_critique", "review.independent_reader", "review.machine_critique", "step:obtain_source_data blocked on data:accession_absent", "step:run_analysis blocked on method:machine_learning_random_forest"], "assays": [{"role": "data_generation", "definition_id": "rna_seq", "definition_version": "99e1195c7d6f", "definition_digest": null, "required_operations": [], "support_state": "partial", "source_hashes": {"assay_lexicon": "99e1195c7d6f"}, "confidence": 0.99, "evidence": ["MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome inf", "ll RNA sequencing (scRNA-seq) captures whole transcriptome information of individual cells. While scRNA"], "order_basis": "paper_text_order:103", "id": "assay:1"}, {"role": "data_generation", "definition_id": "scrna_seq", "definition_version": "99e1195c7d6f", "definition_digest": null, "required_operations": [], "support_state": "unsupported", "source_hashes": {"assay_lexicon": "99e1195c7d6f"}, "confidence": 0.99, "evidence": ["MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome inf", "MOTIVATION: Single-cell RNA sequencing (scRNA-seq) captures whole transcriptome information of", "ptome information of individual cells. While scRNA-seq measures thousands of genes, researchers are", "s how to select those informative genes from scRNA-seq data. Moreover, single-cell targeted gene pr"], "order_basis": "paper_text_order:103", "id": "assay:2"}, {"role": "analysis", "definition_id": "machine_learning", "definition_version": "f2132e96ae6c7a4e", "definition_digest": null, "required_operations": [], "support_state": "partial", "source_hashes": {"dsl_catalogue": "f2132e96ae6c7a4e"}, "confidence": 0.8, "evidence": ["random forest"], "order_basis": "consumes_earlier_output", "id": "assay:3"}]}