Relational Graph Convolutional Network for Robust Mass Spectrum Classification

Supervised machine learning methods have shown impressive performance in interpreting mass signals and automatically segmenting spatially meaningful regions in Mass Spectrometry Imaging (MSI). Such segmentation generates maps that provide researchers with valuable insights into sample composition an...

Full description

Saved in:
Bibliographic Details
Published in:Journal of the American Society for Mass Spectrometry Vol. 36; no. 10; p. 2036
Main Authors: La Rocca, Raphaël, Cioppa, Anthony, Ferrarini, Enrico, Höfte, Monica, Van Droogenbroeck, Marc, De Pauw, Edwin, Eppe, Gauthier, Quinton, Loïc
Format: Journal Article
Language:English
Published: United States 01.10.2025
ISSN:1879-1123, 1879-1123
Online Access:Get more information
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract Supervised machine learning methods have shown impressive performance in interpreting mass signals and automatically segmenting spatially meaningful regions in Mass Spectrometry Imaging (MSI). Such segmentation generates maps that provide researchers with valuable insights into sample composition and serve as a foundation for downstream statistical analyses. However, these models often require data set-specific preprocessing and do not fully exploit the rich mass features available in high-resolution mass spectrometry (HRMS). Unlike low-resolution mass spectrometry, HRMS reveals additional features such as mass defects and repeated mass differences that carry important chemical information. In this work, we propose a novel deep learning architecture based on a Relational Graph Convolutional Network (R-GCN) that captures and leverages those HRMS mass features. Our model explicitly encodes structural features such as mass defects and known mass differences to represent each spectrum as a graph, enabling the learning of associations between chemically related ion families. To the best of our knowledge, no existing deep learning models for MSI classification incorporate this level of chemically informed mass structure. Most existing methods treat spectra as flat vectors or image-like inputs, thereby ignoring the underlying mass relationships. We evaluate our R-GCN approach against several conventional machine learning and deep learning baselines across diverse MSI data sets, demonstrating its robustness to common signal variations (e.g., mass shift, ion loss). Finally, we integrate Class Activation Mapping (CAM) to enhance model interpretability, enabling the identification of ion families that are relevant to specific biological or spatial regions.
AbstractList Supervised machine learning methods have shown impressive performance in interpreting mass signals and automatically segmenting spatially meaningful regions in Mass Spectrometry Imaging (MSI). Such segmentation generates maps that provide researchers with valuable insights into sample composition and serve as a foundation for downstream statistical analyses. However, these models often require data set-specific preprocessing and do not fully exploit the rich mass features available in high-resolution mass spectrometry (HRMS). Unlike low-resolution mass spectrometry, HRMS reveals additional features such as mass defects and repeated mass differences that carry important chemical information. In this work, we propose a novel deep learning architecture based on a Relational Graph Convolutional Network (R-GCN) that captures and leverages those HRMS mass features. Our model explicitly encodes structural features such as mass defects and known mass differences to represent each spectrum as a graph, enabling the learning of associations between chemically related ion families. To the best of our knowledge, no existing deep learning models for MSI classification incorporate this level of chemically informed mass structure. Most existing methods treat spectra as flat vectors or image-like inputs, thereby ignoring the underlying mass relationships. We evaluate our R-GCN approach against several conventional machine learning and deep learning baselines across diverse MSI data sets, demonstrating its robustness to common signal variations (e.g., mass shift, ion loss). Finally, we integrate Class Activation Mapping (CAM) to enhance model interpretability, enabling the identification of ion families that are relevant to specific biological or spatial regions.
Supervised machine learning methods have shown impressive performance in interpreting mass signals and automatically segmenting spatially meaningful regions in Mass Spectrometry Imaging (MSI). Such segmentation generates maps that provide researchers with valuable insights into sample composition and serve as a foundation for downstream statistical analyses. However, these models often require data set-specific preprocessing and do not fully exploit the rich mass features available in high-resolution mass spectrometry (HRMS). Unlike low-resolution mass spectrometry, HRMS reveals additional features such as mass defects and repeated mass differences that carry important chemical information. In this work, we propose a novel deep learning architecture based on a Relational Graph Convolutional Network (R-GCN) that captures and leverages those HRMS mass features. Our model explicitly encodes structural features such as mass defects and known mass differences to represent each spectrum as a graph, enabling the learning of associations between chemically related ion families. To the best of our knowledge, no existing deep learning models for MSI classification incorporate this level of chemically informed mass structure. Most existing methods treat spectra as flat vectors or image-like inputs, thereby ignoring the underlying mass relationships. We evaluate our R-GCN approach against several conventional machine learning and deep learning baselines across diverse MSI data sets, demonstrating its robustness to common signal variations (e.g., mass shift, ion loss). Finally, we integrate Class Activation Mapping (CAM) to enhance model interpretability, enabling the identification of ion families that are relevant to specific biological or spatial regions.Supervised machine learning methods have shown impressive performance in interpreting mass signals and automatically segmenting spatially meaningful regions in Mass Spectrometry Imaging (MSI). Such segmentation generates maps that provide researchers with valuable insights into sample composition and serve as a foundation for downstream statistical analyses. However, these models often require data set-specific preprocessing and do not fully exploit the rich mass features available in high-resolution mass spectrometry (HRMS). Unlike low-resolution mass spectrometry, HRMS reveals additional features such as mass defects and repeated mass differences that carry important chemical information. In this work, we propose a novel deep learning architecture based on a Relational Graph Convolutional Network (R-GCN) that captures and leverages those HRMS mass features. Our model explicitly encodes structural features such as mass defects and known mass differences to represent each spectrum as a graph, enabling the learning of associations between chemically related ion families. To the best of our knowledge, no existing deep learning models for MSI classification incorporate this level of chemically informed mass structure. Most existing methods treat spectra as flat vectors or image-like inputs, thereby ignoring the underlying mass relationships. We evaluate our R-GCN approach against several conventional machine learning and deep learning baselines across diverse MSI data sets, demonstrating its robustness to common signal variations (e.g., mass shift, ion loss). Finally, we integrate Class Activation Mapping (CAM) to enhance model interpretability, enabling the identification of ion families that are relevant to specific biological or spatial regions.
Author De Pauw, Edwin
La Rocca, Raphaël
Quinton, Loïc
Van Droogenbroeck, Marc
Ferrarini, Enrico
Cioppa, Anthony
Höfte, Monica
Eppe, Gauthier
Author_xml – sequence: 1
  givenname: Raphaël
  orcidid: 0000-0003-3388-3041
  surname: La Rocca
  fullname: La Rocca, Raphaël
  organization: Mass Spectrometry Laboratory, MolSys Research Unit, University of Liège, B4000, Liège, Belgium
– sequence: 2
  givenname: Anthony
  surname: Cioppa
  fullname: Cioppa, Anthony
  organization: TELIM, Montefiore Institute, University of Liège, B4000, Liège, Belgium
– sequence: 3
  givenname: Enrico
  surname: Ferrarini
  fullname: Ferrarini, Enrico
  organization: Department of Plants and Crops, Faculty of Bioscience Engineering, Ghent University, B9000, Ghent, Belgium
– sequence: 4
  givenname: Monica
  orcidid: 0000-0002-0850-3249
  surname: Höfte
  fullname: Höfte, Monica
  organization: Department of Plants and Crops, Faculty of Bioscience Engineering, Ghent University, B9000, Ghent, Belgium
– sequence: 5
  givenname: Marc
  surname: Van Droogenbroeck
  fullname: Van Droogenbroeck, Marc
  organization: TELIM, Montefiore Institute, University of Liège, B4000, Liège, Belgium
– sequence: 6
  givenname: Edwin
  orcidid: 0000-0003-3475-1315
  surname: De Pauw
  fullname: De Pauw, Edwin
  organization: Mass Spectrometry Laboratory, MolSys Research Unit, University of Liège, B4000, Liège, Belgium
– sequence: 7
  givenname: Gauthier
  orcidid: 0000-0002-4821-3115
  surname: Eppe
  fullname: Eppe, Gauthier
  organization: Mass Spectrometry Laboratory, MolSys Research Unit, University of Liège, B4000, Liège, Belgium
– sequence: 8
  givenname: Loïc
  orcidid: 0000-0001-8153-9590
  surname: Quinton
  fullname: Quinton, Loïc
  organization: Mass Spectrometry Laboratory, MolSys Research Unit, University of Liège, B4000, Liège, Belgium
BackLink https://www.ncbi.nlm.nih.gov/pubmed/40888691$$D View this record in MEDLINE/PubMed
BookMark eNpNkM1LwzAYxoNM3IcevUqOXjrfJEuaHqXoFOaEuXtJ0hQ706YmreJ_79AKnp4PfjyHZ44mrW8tQpcElgQouTmo2MQlNwDA-QmaEZlmCSGUTf75KZrHeAAgKWTpGZquQEopMjJD2511qq99qxxeB9W94ty3H94NY7e1_acPb7jyAe-8HmKPn1SM-KWzpg9Dg3N3jHVVm5-Vc3RaKRftxagLtL-_2-cPyeZ5_ZjfbhLFhOwToowVhptSgWU8gzKTKishrSi1gmtdCkKU0FqqVFNiGDUMaFVpwzWF0tIFuv6d7YJ_H2zsi6aOxjqnWuuHWDC6EiyVEugRvRrRQTe2LLpQNyp8FX8X0G96fGGB
ContentType Journal Article
DBID NPM
7X8
DOI 10.1021/jasms.5c00055
DatabaseName PubMed
MEDLINE - Academic
DatabaseTitle PubMed
MEDLINE - Academic
DatabaseTitleList PubMed
MEDLINE - Academic
Database_xml – sequence: 1
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 2
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod no_fulltext_linktorsrc
Discipline Chemistry
EISSN 1879-1123
ExternalDocumentID 40888691
Genre Journal Article
GroupedDBID ---
06D
0VY
199
1B1
1~5
203
2KG
2KM
2LR
30V
4.4
408
40E
4G.
5C9
5GY
67N
71M
8UJ
95.
96X
AABHQ
AAJKR
AAKDD
AARTL
AATVU
AAWCG
AAYIU
AAYQN
AAZMS
ABBLG
ABFTV
ABHLI
ABJNI
ABJOX
ABKCH
ABLBI
ABQBU
ABQRX
ABTHY
ABTMW
ABXPI
ACGFO
ACGFS
ACHXU
ACKNC
ACMLO
ACREN
ACS
ADBBV
ADEZE
ADHHG
ADHIR
ADKPE
ADURQ
ADYOE
ADZKW
AEGAL
AEGNC
AEJHL
AEJRE
AENEX
AEOHA
AEPYU
AETCA
AEXYK
AFBBN
AFWTZ
AFYQB
AFZKB
AGAYW
AGDGC
AGQMX
AGWZB
AGYKE
AHAVH
AHBYD
AHGAQ
AHKAY
AHYZX
AIIXL
AITGF
AJRNO
AKRWK
ALMA_UNASSIGNED_HOLDINGS
AMKLP
AMTXH
AMYQR
ANMIH
ARMRJ
ASPBG
AVWKF
AXYYD
AYJHY
AZFZN
BAANH
BGLVJ
CS3
CSCUP
CUPRZ
DIK
DU5
DX2
E3Z
EBS
EIOEI
EN4
ESBYG
F5P
FDB
FFXSO
FRRFC
G-Q
GGK
GGRSB
GJIRD
GQ7
HF~
HMJXF
HRMNR
HZ~
IHE
ITM
IXB
JBSCW
KOV
NPM
NQ-
NQJWS
O-L
O93
O9I
O9J
OZT
P2P
PF-
R9I
RPZ
RSV
S16
S1Z
S27
S3A
S3B
SBL
SDG
SEW
SHX
SISQX
SOJ
SPISZ
STPWE
SZN
T13
TR2
TSG
U2A
UG4
VC2
VF5
VG9
WK8
YK3
~HD
7X8
ID FETCH-LOGICAL-a368t-1ace6c5cda0e3590d98a9d07f22e65bbd611a6bb8a7b21c32c302ffbc5b20de2
IEDL.DBID 7X8
ISICitedReferencesCount 0
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001561628500001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1879-1123
IngestDate Thu Oct 02 21:42:46 EDT 2025
Thu Oct 02 04:49:19 EDT 2025
IsDoiOpenAccess false
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 10
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-a368t-1ace6c5cda0e3590d98a9d07f22e65bbd611a6bb8a7b21c32c302ffbc5b20de2
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ORCID 0000-0002-4821-3115
0000-0001-8153-9590
0000-0002-0850-3249
0000-0003-3475-1315
0000-0003-3388-3041
OpenAccessLink https://orbi.uliege.be/handle/2268/335949
PMID 40888691
PQID 3246378802
PQPubID 23479
ParticipantIDs proquest_miscellaneous_3246378802
pubmed_primary_40888691
PublicationCentury 2000
PublicationDate 2025-10-01
PublicationDateYYYYMMDD 2025-10-01
PublicationDate_xml – month: 10
  year: 2025
  text: 2025-10-01
  day: 01
PublicationDecade 2020
PublicationPlace United States
PublicationPlace_xml – name: United States
PublicationTitle Journal of the American Society for Mass Spectrometry
PublicationTitleAlternate J Am Soc Mass Spectrom
PublicationYear 2025
SSID ssj0017097
Score 2.4642553
Snippet Supervised machine learning methods have shown impressive performance in interpreting mass signals and automatically segmenting spatially meaningful regions in...
SourceID proquest
pubmed
SourceType Aggregation Database
Index Database
StartPage 2036
Title Relational Graph Convolutional Network for Robust Mass Spectrum Classification
URI https://www.ncbi.nlm.nih.gov/pubmed/40888691
https://www.proquest.com/docview/3246378802
Volume 36
WOSCitedRecordID wos001561628500001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1bS8MwFA7qBH3xfpk3Ivgal6ZN0zyJDKcPrgwZsreSW0Fh7Vy7_X6TtMUnQfCl0NKU5nByztecr-cD4I4YyUxCBdKKYhTZDw7EeZ4gzUnOfOXH99J7f2VpmsxmfNJuuFUtrbKLiT5Q61K5PfKBTfyx632OycPiCznVKFddbSU0NkEvtFDGUbrY7KeKwHAjrpIwjiyuCNsemzatDT5FNa_uqXKYhf6OLn2WGe3_9_0OwF6LL-Fj4xCHYMMUR2Bn2Mm6HYO0o7_Zu55dt2o4LIt164D2WtrwwqEFs_CtlKuqhmMLsKETqq-Xqzn0MpqOYOSfcgKmo6fp8AW1ogpIhHFSo0AoEyuqtMAmpBxrngiuMcsJMTGVUsdBIGIpE8EkCVRIVIhJnktFJcHakFOwVZSFOQdQaB6KCKvAorhIGMylEO6_VnsuAyZxH9x2lsrsHF0hQhSmXFXZj6364Kwxd7ZommtkkQ17ScyDiz-MvgS7xMnxem7dFejldsWaa7Ct1vVHtbzxzmCP6WT8DciuwHY
linkProvider ProQuest
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Relational+Graph+Convolutional+Network+for+Robust+Mass+Spectrum+Classification&rft.jtitle=Journal+of+the+American+Society+for+Mass+Spectrometry&rft.au=La+Rocca%2C+Rapha%C3%ABl&rft.au=Cioppa%2C+Anthony&rft.au=Ferrarini%2C+Enrico&rft.au=H%C3%B6fte%2C+Monica&rft.date=2025-10-01&rft.issn=1879-1123&rft.eissn=1879-1123&rft_id=info:doi/10.1021%2Fjasms.5c00055&rft.externalDBID=NO_FULL_TEXT
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1879-1123&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1879-1123&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1879-1123&client=summon