Pretraining Generative Flow Networks with Inexpensive Rewards for Molecular Graph Generation Article Swipe

PDF

Mohit Pandey , Gopeshh Subbaraj , Artem Cherkasov , Martin Ester , Emmanuel Bengio ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2503.06337

Generative Flow Networks (GFlowNets) have recently emerged as a suitable framework for generating diverse and high-quality molecular structures by learning from rewards treated as unnormalized distributions. Previous works in this framework often restrict exploration by using predefined molecular fragments as building blocks, limiting the chemical space that can be accessed. In this work, we introduce Atomic GFlowNets (A-GFNs), a foundational generative model leveraging individual atoms as building blocks to explore drug-like chemical space more comprehensively. We propose an unsupervised pre-training approach using drug-like molecule datasets, which teaches A-GFNs about inexpensive yet informative molecular descriptors such as drug-likeliness, topological polar surface area, and synthetic accessibility scores. These properties serve as proxy rewards, guiding A-GFNs towards regions of chemical space that exhibit desirable pharmacological properties. We further implement a goal-conditioned finetuning process, which adapts A-GFNs to optimize for specific target properties. In this work, we pretrain A-GFN on a subset of ZINC dataset, and by employing robust evaluation metrics we show the effectiveness of our approach when compared to other relevant baseline methods for a wide range of drug design tasks. The code is accessible at https://github.com/diamondspark/AGFN.

Related Topics

Generative Grammar

Computer Science

Artificial Intelligence

Theoretical Computer Science

Mathematics

Geometry

Concepts

Generative grammar Graph Computer science Flow (mathematics) Control flow graph Artificial intelligence Theoretical computer science Mathematics Geometry

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2503.06337
PDF: https://arxiv.org/pdf/2503.06337
OA Status: green
Related Works: 10
OpenAlex ID: https://openalex.org/W4408973838

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4408973838

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2503.06337

Digital Object Identifier
Title: Pretraining Generative Flow Networks with Inexpensive Rewards for Molecular Graph Generation

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-03-08

Full publication date if available
Authors: Mohit Pandey, Gopeshh Subbaraj, Artem Cherkasov, Martin Ester, Emmanuel Bengio

List of authors in order
Landing page: https://arxiv.org/abs/2503.06337

Publisher landing page
PDF URL: https://arxiv.org/pdf/2503.06337

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: green

Open access status per OpenAlex
OA URL: https://arxiv.org/pdf/2503.06337

Direct OA link when available
Concepts: Generative grammar, Graph, Computer science, Flow (mathematics), Control flow graph, Artificial intelligence, Theoretical computer science, Mathematics, Geometry

Top concepts (fields/topics) attached by OpenAlex
Cited by: 0

Total citation count in OpenAlex
Related works (count): 10

Other works algorithmically related by OpenAlex

Full payload

id	https://openalex.org/W4408973838
doi	https://doi.org/10.48550/arxiv.2503.06337
ids.doi	https://doi.org/10.48550/arxiv.2503.06337
ids.openalex	https://openalex.org/W4408973838
fwci
type	preprint
title	Pretraining Generative Flow Networks with Inexpensive Rewards for Molecular Graph Generation
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T11407
topics[0].field.id	https://openalex.org/fields/22
topics[0].field.display_name	Engineering
topics[0].score	0.9944000244140625
topics[0].domain.id	https://openalex.org/domains/3
topics[0].domain.display_name	Physical Sciences
topics[0].subfield.id	https://openalex.org/subfields/2204
topics[0].subfield.display_name	Biomedical Engineering
topics[0].display_name	Innovative Microfluidic and Catalytic Techniques Innovation
topics[1].id	https://openalex.org/T11948
topics[1].field.id	https://openalex.org/fields/25
topics[1].field.display_name	Materials Science
topics[1].score	0.9337000250816345
topics[1].domain.id	https://openalex.org/domains/3
topics[1].domain.display_name	Physical Sciences
topics[1].subfield.id	https://openalex.org/subfields/2505
topics[1].subfield.display_name	Materials Chemistry
topics[1].display_name	Machine Learning in Materials Science
topics[2].id	https://openalex.org/T10211
topics[2].field.id	https://openalex.org/fields/17
topics[2].field.display_name	Computer Science
topics[2].score	0.9279999732971191
topics[2].domain.id	https://openalex.org/domains/3
topics[2].domain.display_name	Physical Sciences
topics[2].subfield.id	https://openalex.org/subfields/1703
topics[2].subfield.display_name	Computational Theory and Mathematics
topics[2].display_name	Computational Drug Discovery Methods
is_xpac	False
apc_list
apc_paid
concepts[0].id	https://openalex.org/C39890363
concepts[0].level	2
concepts[0].score	0.8049079179763794
concepts[0].wikidata	https://www.wikidata.org/wiki/Q36108
concepts[0].display_name	Generative grammar
concepts[1].id	https://openalex.org/C132525143
concepts[1].level	2
concepts[1].score	0.575545608997345
concepts[1].wikidata	https://www.wikidata.org/wiki/Q141488
concepts[1].display_name	Graph
concepts[2].id	https://openalex.org/C41008148
concepts[2].level	0
concepts[2].score	0.5203951597213745
concepts[2].wikidata	https://www.wikidata.org/wiki/Q21198
concepts[2].display_name	Computer science
concepts[3].id	https://openalex.org/C38349280
concepts[3].level	2
concepts[3].score	0.47800102829933167
concepts[3].wikidata	https://www.wikidata.org/wiki/Q1434290
concepts[3].display_name	Flow (mathematics)
concepts[4].id	https://openalex.org/C27458966
concepts[4].level	2
concepts[4].score	0.4164818823337555
concepts[4].wikidata	https://www.wikidata.org/wiki/Q1187693
concepts[4].display_name	Control flow graph
concepts[5].id	https://openalex.org/C154945302
concepts[5].level	1
concepts[5].score	0.32617396116256714
concepts[5].wikidata	https://www.wikidata.org/wiki/Q11660
concepts[5].display_name	Artificial intelligence
concepts[6].id	https://openalex.org/C80444323
concepts[6].level	1
concepts[6].score	0.29187750816345215
concepts[6].wikidata	https://www.wikidata.org/wiki/Q2878974
concepts[6].display_name	Theoretical computer science
concepts[7].id	https://openalex.org/C33923547
concepts[7].level	0
concepts[7].score	0.2306532859802246
concepts[7].wikidata	https://www.wikidata.org/wiki/Q395
concepts[7].display_name	Mathematics
concepts[8].id	https://openalex.org/C2524010
concepts[8].level	1
concepts[8].score	0.0
concepts[8].wikidata	https://www.wikidata.org/wiki/Q8087
concepts[8].display_name	Geometry
keywords[0].id	https://openalex.org/keywords/generative-grammar
keywords[0].score	0.8049079179763794
keywords[0].display_name	Generative grammar
keywords[1].id	https://openalex.org/keywords/graph
keywords[1].score	0.575545608997345
keywords[1].display_name	Graph
keywords[2].id	https://openalex.org/keywords/computer-science
keywords[2].score	0.5203951597213745
keywords[2].display_name	Computer science
keywords[3].id	https://openalex.org/keywords/flow
keywords[3].score	0.47800102829933167
keywords[3].display_name	Flow (mathematics)
keywords[4].id	https://openalex.org/keywords/control-flow-graph
keywords[4].score	0.4164818823337555
keywords[4].display_name	Control flow graph
keywords[5].id	https://openalex.org/keywords/artificial-intelligence
keywords[5].score	0.32617396116256714
keywords[5].display_name	Artificial intelligence
keywords[6].id	https://openalex.org/keywords/theoretical-computer-science
keywords[6].score	0.29187750816345215
keywords[6].display_name	Theoretical computer science
keywords[7].id	https://openalex.org/keywords/mathematics
keywords[7].score	0.2306532859802246
keywords[7].display_name	Mathematics
language	en
locations[0].id	pmh:oai:arXiv.org:2503.06337
locations[0].is_oa	True
locations[0].source.id	https://openalex.org/S4306400194
locations[0].source.issn
locations[0].source.type	repository
locations[0].source.is_oa	True
locations[0].source.issn_l
locations[0].source.is_core	False
locations[0].source.is_in_doaj	False
locations[0].source.display_name	arXiv (Cornell University)
locations[0].source.host_organization	https://openalex.org/I205783295
locations[0].source.host_organization_name	Cornell University
locations[0].source.host_organization_lineage	https://openalex.org/I205783295
locations[0].license
locations[0].pdf_url	https://arxiv.org/pdf/2503.06337
locations[0].version	submittedVersion
locations[0].raw_type	text
locations[0].license_id
locations[0].is_accepted	False
locations[0].is_published	False
locations[0].raw_source_name
locations[0].landing_page_url	http://arxiv.org/abs/2503.06337
locations[1].id	doi:10.48550/arxiv.2503.06337
locations[1].is_oa	True
locations[1].source.id	https://openalex.org/S4306400194
locations[1].source.issn
locations[1].source.type	repository
locations[1].source.is_oa	True
locations[1].source.issn_l
locations[1].source.is_core	False
locations[1].source.is_in_doaj	False
locations[1].source.display_name	arXiv (Cornell University)
locations[1].source.host_organization	https://openalex.org/I205783295
locations[1].source.host_organization_name	Cornell University
locations[1].source.host_organization_lineage	https://openalex.org/I205783295
locations[1].license	cc-by
locations[1].pdf_url
locations[1].version
locations[1].raw_type	article
locations[1].license_id	https://openalex.org/licenses/cc-by
locations[1].is_accepted	False
locations[1].is_published
locations[1].raw_source_name
locations[1].landing_page_url	https://doi.org/10.48550/arxiv.2503.06337
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5014143640
authorships[0].author.orcid	https://orcid.org/0000-0002-2562-7155
authorships[0].author.display_name	Mohit Pandey
authorships[0].author_position	first
authorships[0].raw_author_name	Pandey, Mohit
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5114369609
authorships[1].author.orcid
authorships[1].author.display_name	Gopeshh Subbaraj
authorships[1].author_position	middle
authorships[1].raw_author_name	Subbaraj, Gopeshh
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5070580886
authorships[2].author.orcid	https://orcid.org/0000-0002-1599-1439
authorships[2].author.display_name	Artem Cherkasov
authorships[2].author_position	middle
authorships[2].raw_author_name	Cherkasov, Artem
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5018267399
authorships[3].author.orcid	https://orcid.org/0000-0001-7732-2815
authorships[3].author.display_name	Martin Ester
authorships[3].author_position	middle
authorships[3].raw_author_name	Ester, Martin
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5114369610
authorships[4].author.orcid
authorships[4].author.display_name	Emmanuel Bengio
authorships[4].author_position	last
authorships[4].raw_author_name	Bengio, Emmanuel
authorships[4].is_corresponding	False
has_content.pdf	False
has_content.grobid_xml	False
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://arxiv.org/pdf/2503.06337
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	Pretraining Generative Flow Networks with Inexpensive Rewards for Molecular Graph Generation
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T11407
primary_topic.field.id	https://openalex.org/fields/22
primary_topic.field.display_name	Engineering
primary_topic.score	0.9944000244140625
primary_topic.domain.id	https://openalex.org/domains/3
primary_topic.domain.display_name	Physical Sciences
primary_topic.subfield.id	https://openalex.org/subfields/2204
primary_topic.subfield.display_name	Biomedical Engineering
primary_topic.display_name	Innovative Microfluidic and Catalytic Techniques Innovation
related_works	https://openalex.org/W2380075625, https://openalex.org/W4390718435, https://openalex.org/W4390549206, https://openalex.org/W3137171911, https://openalex.org/W4379540039, https://openalex.org/W4237784285, https://openalex.org/W2374712251, https://openalex.org/W4383031710, https://openalex.org/W3211753092, https://openalex.org/W2386000789
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2503.06337
best_oa_location.is_oa	True
best_oa_location.source.id	https://openalex.org/S4306400194
best_oa_location.source.issn
best_oa_location.source.type	repository
best_oa_location.source.is_oa	True
best_oa_location.source.issn_l
best_oa_location.source.is_core	False
best_oa_location.source.is_in_doaj	False
best_oa_location.source.display_name	arXiv (Cornell University)
best_oa_location.source.host_organization	https://openalex.org/I205783295
best_oa_location.source.host_organization_name	Cornell University
best_oa_location.source.host_organization_lineage	https://openalex.org/I205783295
best_oa_location.license
best_oa_location.pdf_url	https://arxiv.org/pdf/2503.06337
best_oa_location.version	submittedVersion
best_oa_location.raw_type	text
best_oa_location.license_id
best_oa_location.is_accepted	False
best_oa_location.is_published	False
best_oa_location.raw_source_name
best_oa_location.landing_page_url	http://arxiv.org/abs/2503.06337
primary_location.id	pmh:oai:arXiv.org:2503.06337
primary_location.is_oa	True
primary_location.source.id	https://openalex.org/S4306400194
primary_location.source.issn
primary_location.source.type	repository
primary_location.source.is_oa	True
primary_location.source.issn_l
primary_location.source.is_core	False
primary_location.source.is_in_doaj	False
primary_location.source.display_name	arXiv (Cornell University)
primary_location.source.host_organization	https://openalex.org/I205783295
primary_location.source.host_organization_name	Cornell University
primary_location.source.host_organization_lineage	https://openalex.org/I205783295
primary_location.license
primary_location.pdf_url	https://arxiv.org/pdf/2503.06337
primary_location.version	submittedVersion
primary_location.raw_type	text
primary_location.license_id
primary_location.is_accepted	False
primary_location.is_published	False
primary_location.raw_source_name
primary_location.landing_page_url	http://arxiv.org/abs/2503.06337
publication_date	2025-03-08
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	8, 58, 126, 146, 172
abstract_inverted_index.In	50, 139
abstract_inverted_index.We	75, 123
abstract_inverted_index.an	77
abstract_inverted_index.as	7, 23, 39, 65, 95, 108
abstract_inverted_index.at	183
abstract_inverted_index.be	48
abstract_inverted_index.by	18, 34, 152
abstract_inverted_index.in	28
abstract_inverted_index.is	181
abstract_inverted_index.of	115, 148, 161, 175
abstract_inverted_index.on	145
abstract_inverted_index.to	68, 133, 166
abstract_inverted_index.we	53, 142, 157
abstract_inverted_index.The	179
abstract_inverted_index.and	14, 101, 151
abstract_inverted_index.can	47
abstract_inverted_index.for	11, 135, 171
abstract_inverted_index.our	162
abstract_inverted_index.the	43, 159
abstract_inverted_index.yet	90
abstract_inverted_index.Flow	1
abstract_inverted_index.ZINC	149
abstract_inverted_index.code	180
abstract_inverted_index.drug	176
abstract_inverted_index.from	20
abstract_inverted_index.have	4
abstract_inverted_index.more	73
abstract_inverted_index.show	158
abstract_inverted_index.such	94
abstract_inverted_index.that	46, 118
abstract_inverted_index.this	29, 51, 140
abstract_inverted_index.when	164
abstract_inverted_index.wide	173
abstract_inverted_index.A-GFN	144
abstract_inverted_index.These	105
abstract_inverted_index.about	88
abstract_inverted_index.area,	100
abstract_inverted_index.atoms	64
abstract_inverted_index.model	61
abstract_inverted_index.often	31
abstract_inverted_index.other	167
abstract_inverted_index.polar	98
abstract_inverted_index.proxy	109
abstract_inverted_index.range	174
abstract_inverted_index.serve	107
abstract_inverted_index.space	45, 72, 117
abstract_inverted_index.using	35, 81
abstract_inverted_index.which	85, 130
abstract_inverted_index.work,	52, 141
abstract_inverted_index.works	27
abstract_inverted_index.A-GFNs	87, 112, 132
abstract_inverted_index.Atomic	55
abstract_inverted_index.adapts	131
abstract_inverted_index.blocks	67
abstract_inverted_index.design	177
abstract_inverted_index.robust	154
abstract_inverted_index.subset	147
abstract_inverted_index.target	137
abstract_inverted_index.tasks.	178
abstract_inverted_index.blocks,	41
abstract_inverted_index.diverse	13
abstract_inverted_index.emerged	6
abstract_inverted_index.exhibit	119
abstract_inverted_index.explore	69
abstract_inverted_index.further	124
abstract_inverted_index.guiding	111
abstract_inverted_index.methods	170
abstract_inverted_index.metrics	156
abstract_inverted_index.propose	76
abstract_inverted_index.regions	114
abstract_inverted_index.rewards	21
abstract_inverted_index.scores.	104
abstract_inverted_index.surface	99
abstract_inverted_index.teaches	86
abstract_inverted_index.towards	113
abstract_inverted_index.treated	22
abstract_inverted_index.Networks	2
abstract_inverted_index.Previous	26
abstract_inverted_index.approach	80, 163
abstract_inverted_index.baseline	169
abstract_inverted_index.building	40, 66
abstract_inverted_index.chemical	44, 71, 116
abstract_inverted_index.compared	165
abstract_inverted_index.dataset,	150
abstract_inverted_index.learning	19
abstract_inverted_index.limiting	42
abstract_inverted_index.molecule	83
abstract_inverted_index.optimize	134
abstract_inverted_index.pretrain	143
abstract_inverted_index.process,	129
abstract_inverted_index.recently	5
abstract_inverted_index.relevant	168
abstract_inverted_index.restrict	32
abstract_inverted_index.rewards,	110
abstract_inverted_index.specific	136
abstract_inverted_index.suitable	9
abstract_inverted_index.(A-GFNs),	57
abstract_inverted_index.GFlowNets	56
abstract_inverted_index.accessed.	49
abstract_inverted_index.datasets,	84
abstract_inverted_index.desirable	120
abstract_inverted_index.drug-like	70, 82
abstract_inverted_index.employing	153
abstract_inverted_index.fragments	38
abstract_inverted_index.framework	10, 30
abstract_inverted_index.implement	125
abstract_inverted_index.introduce	54
abstract_inverted_index.molecular	16, 37, 92
abstract_inverted_index.synthetic	102
abstract_inverted_index.Generative	0
abstract_inverted_index.accessible	182
abstract_inverted_index.evaluation	155
abstract_inverted_index.finetuning	128
abstract_inverted_index.generating	12
abstract_inverted_index.generative	60
abstract_inverted_index.individual	63
abstract_inverted_index.leveraging	62
abstract_inverted_index.predefined	36
abstract_inverted_index.properties	106
abstract_inverted_index.structures	17
abstract_inverted_index.(GFlowNets)	3
abstract_inverted_index.descriptors	93
abstract_inverted_index.exploration	33
abstract_inverted_index.inexpensive	89
abstract_inverted_index.informative	91
abstract_inverted_index.properties.	122, 138
abstract_inverted_index.topological	97
abstract_inverted_index.foundational	59
abstract_inverted_index.high-quality	15
abstract_inverted_index.pre-training	79
abstract_inverted_index.unnormalized	24
abstract_inverted_index.unsupervised	78
abstract_inverted_index.accessibility	103
abstract_inverted_index.effectiveness	160
abstract_inverted_index.distributions.	25
abstract_inverted_index.pharmacological	121
abstract_inverted_index.comprehensively.	74
abstract_inverted_index.drug-likeliness,	96
abstract_inverted_index.goal-conditioned	127
abstract_inverted_index.https://github.com/diamondspark/AGFN.	184
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	5
citation_normalized_percentile