Home Big Data Mining and Analytics Article
PDF (3.8 MB)
Cite
EndNote(RIS) BibTeX
Collect
Submit Manuscript
Show Outline
Figures (4)

Tables (3)
Table 1
Table 2
Table 3
Open Access

Exploring Trial-and-Error in Deep Learning: Initial Application to Isotope Detection in Mass Spectrometry

School of Computer Science and Technology, Shandong University, Qingdao 266237, China
Key Lab of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China

Show Author Information

Abstract

Mass spectrometry plays a crucial role in biomedicine by detecting isotopes, contributing significantly to research, diagnostics, and therapy development. This study introduces IsoFusion, a deep learning model designed to address isotope detection in raw mass spectra. Rather than directly applying convolutional layers to all signal and noise peaks, IsoFusion employs a trial-and-error strategy. First, it investigates all potential charge states (trials) and collects signal peaks around expected m/z values for each trial. Then, convolutional layers extract features from each trial, which are fused to identify the correct one. Finally, the reparameterization trick predicts isotope features based on this correct trial. A key advantage of IsoFusion is shared model parameters across all trials, enhancing feature learning for less common charge states using data from prevalent ones. Our results show a significant accuracy improvement for charge state 5, reaching 99.42%, compared to DeepIso’s 43.36%. Moreover, IsoFusion achieves a 97.33% detection accuracy for isotopes, with 2.4% of detected isotopes previously unidentified by four commonly used methods.

Keywords

Liquid Chromatography-Mass Spectrometry (LC-MS)isotope detectionretention time predictioncharge state predictiondeep learning

References

[1]

V. Kumar, J. D. Lee, R. J. Clark, and T. M. Woodruff, Development and validation of a LC-MS/MS assay for pharmacokinetic studies of complement C5a receptor antagonists PMX53 and PMX205 in mice, Sci. Rep., vol. 8, no. 1, p. 8101, 2018.

Crossref Google Scholar
[2]

L. Chang, J. Ni, J. Beretov, V. C. Wasinger, J. Hao, J. Bucci, D. Malouf, D. Gillatt, P. H. Graham, and Y. Li, Identification of protein biomarkers and signaling pathways associated with prostate cancer radioresistance using label-free LC-MS/MS proteomic approach, Sci. Rep., vol. 7, no. 1, p. 41834, 2017.

Crossref Google Scholar
[3]

D. Chen, X. Su, N. Wang, Y. Li, H. Yin, L. Li, and L. Li, Chemical isotope labeling LC-MS for monitoring disease progression and treatment in animal models: Plasma metabolomics study of osteoarthritis rat model, Sci. Rep., vol. 7, no. 1, p. 40543, 2017.

Crossref Google Scholar
[4]

F. Becher, J. Ciccolini, D.-C. Imbs, C. Marin, C. Fournel, C. Dupuis, N. Fakhry, B. Pourroy, A. Ghettas, A. Pruvost, et al., A simple and rapid LC-MS/MS method for therapeutic drug monitoring of cetuximab: A GPCO-UNICANCER proof of concept study in head-and-neck cancer patients, Sci. Rep., vol. 7, no. 1, p. 2714, 2017.

Crossref Google Scholar
[5]

Y. Tang, F. Tang, Y. Yang, L. Zhao, H. Zhou, J. Dong, and W. Huang, Real-time analysis on drug-antibody ratio of antibody-drug conjugates for synthesis, process optimization, and quality control, Sci. Rep., vol. 7, no. 1, p. 7763, 2017.

Crossref Google Scholar
[6]

J. Cox and M. Mann, MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification, Nat. Biotechnol., vol. 26, p. 1367–1372, 2008.

Crossref Google Scholar
[7]

P. M. Palagi, D. Walther, M. Quadroni, S. Catherinet, J. Burgess, C. G. Zimmermann-Ivol, J.-C. Sanchez, P.-A. Binz, D. F. Hochstrasser, and R. D. Appel, MSight: An image analysis software for liquid chromatography-mass spectrometry, Proteomics, vol. 5, no. 9, pp. 2381–2384, 2005.

Crossref Google Scholar
[8]

R. Tautenhahn, C. Böttcher, and S. Neumann, Highly sensitive feature detection for high resolution LC/MS, BMC Bioinform., vol. 9, no. 1, p. 504, 2008.

Crossref Google Scholar
[9]

M. Sturm, A. Bertsch, C. Gröpl, A. Hildebrandt, R. Hussong, E. Lange, N. Pfeifer, O. Schulz-Trieglaff, A. Zerck, K. Reinert, et al., OpenMS–An Open-source software framework for mass spectrometry, BMC Bioinform., vol. 9, no. 1, p. 163, 2008.

Crossref Google Scholar
[10]

C. J. Conley, R. Smith, R. J. O. Torgrip, R. M. Taylor, R. Tautenhahn, and J. T. Prince, Massifquant: Open-source Kalman filter-based XC-MS isotope trace feature detection, Bioinformatics, vol. 30, no. 18, pp. 2636–2643, 2014.

Crossref Google Scholar
[11]

K. Aoshima, K. Takahashi, M. Ikawa, T. Kimura, M. Fukuda, S. Tanaka, H. E. Parry, Y. Fujita, A. C. Yoshizawa, S.-I. Utsunomiya, et al., A simple peak detection and label-free quantitation algorithm for chromatography-mass spectrometry, BMC Bioinform., vol. 15, no. 1, p. 376, 2014.

Crossref Google Scholar
[12]

J. Teleman, A. Chawade, M. Sandin, F. Levander, and J. Malmström, Dinosaur: A refined open-source peptide MS feature detector, J. Proteome Res., vol. 15, no. 7, pp. 2143–2151, 2016.

Crossref Google Scholar
[13]

H. Weisser and J. S. Choudhary, Targeted feature detection for data-dependent shotgun proteomics, J. Proteome Res., vol. 16, no. 8, pp. 2964–2974, 2017.

Crossref Google Scholar
[14]
X. Zeng and B. Ma, MSTracer: A machine learning software tool for peptide feature detection from liquid chromatography-mass spectrometry data, J. Proteome Res., vol. 20, no. 7, pp. 3455–3462, 2021.
Crossref
[15]

Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, Gradient-based learning applied to document recognition, Proc. IEEE, vol. 86, no. 11, pp. 2278–2324, 1998.

Crossref Google Scholar
[16]
J. J. Hopfield, Neural networks and physical systems with emergent collective computational abilities, in Proceedings of the National Academy of Sciences of the Unite State of America, doi: 10.1073/pnas.79.8.2554.
Crossref
[17]

F. T. Zohora, M. Z. Rahman, N. H. Tran, L. Xin, B. Shan, and M. Li, DeepIso: A deep learning model for peptide feature detection from LC-MS map, Sci. Rep., vol. 9, no. 1, p. 17168, 2019.

Crossref Google Scholar
[18]
R. Q. Charles, S. Hao, K. Mo, and L. J. Guibas, PointNet: deep learning on point sets for 3D classification and segmentation, in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017, pp. 77–85.
Crossref
[19]
J. Fu, J. Liu, H. Tian, Y. Li, Y. Bao, Z. Fang, and H. Lu, Dual attention network for scene segmentation, in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA, 2019, pp. 3141–3149.
Crossref
[20]

F. T. Zohora, M. Z. Rahman, N. H. Tran, L. Xin, B. Shan, and M. Li, Deep neural network for detecting arbitrary precision peptide features through attention based segmentation, Sci. Rep., vol. 11, no. 1, p. 18249, 2021.

Crossref Google Scholar
[21]

A. R. Basharat, Y. Zang, L. Sun, and X. Liu, TopFD: A proteoform feature detection tool for top-down proteomics, Anal. Chem., vol. 95, no. 21, pp. 8189–8196, 2023.

Crossref Google Scholar
[22]

H. Steen and M. Mann, The abc’s (and xyz’s) of peptide sequencing, Nat. Rev. Mol. Cell Biol., vol. 5, pp. 699–711, 2004.

Crossref Google Scholar
[23]
D. P. Kingma and M. Welling, Auto-encoding variational Bayes, arXiv preprint arXiv: 1312.6114, 2013.
[24]
E. Jang, S. Gu, and B. Poole, Categorical reparameterization with gumbel-softmax, arXiv preprint arXiv: 1611.01144, 2016.
[25]
A. van den Oord, O. Vinyals, and K. Kavukcuoglu, Neural discrete representation learning, arXiv preprint arXiv: 1711.00937, 2017.
[26]

M. C. Chambers, B. MacLean, R. Burke, D. Amodei, D. L. Ruderman, S. Neumann, L. Gatto, B. Fischer, B. Pratt, J. Egertson, et al., A cross-platform toolkit for mass spectrometry and proteomics, Nat. Biotechnol., vol. 30, no. 10, pp. 918–920, 2012.

Crossref Google Scholar
[27]

A. Chawade, M. Sandin, J. Teleman, J. Malmström, and F. Levander, Data processing has major impact on the outcome of quantitative label-free LC-MS analysis, J. Proteome Res., vol. 14, no. 2, pp. 676–687, 2015.

Crossref Google Scholar
[28]
I. Loshchilov and F. Hutter, Decoupled weight decay regularization, arXiv preprint arXiv: 1711.05101, 2017.
[29]

D. N. Perkins, D. J. C. Pappin, D. M. Creasy, and J. S. Cottrell, Probability-based protein identification by searching sequence databases using mass spectrometry data, 3.0.CO;2-2">Electrophoresis, vol. 20, no. 18, pp. 3551–3567, 1999.

Crossref Google Scholar
[30]
B. Ma, K. Zhang, C. Hendrie, C. Liang, M. Li, A. Doherty-Kirby, and G. Lajoie, PEAKS: Powerful software for peptide de novo sequencing by tandem mass spectrometry, Rapid Commun. Mass Spectrom., vol. 17, no. 20, pp. 2337–2342, 2003.
Crossref
Big Data Mining and Analytics
Volume 7 Issue 4,
December 2024
Pages 1251-1261
DOI: 10.26599/BDMA.2024.9020059
Cite this article:
Jiao Q, Wang Y, Wang Y, et al. Exploring Trial-and-Error in Deep Learning: Initial Application to Isotope Detection in Mass Spectrometry. Big Data Mining and Analytics, 2024, 7(4): 1251-1261. https://doi.org/10.26599/BDMA.2024.9020059
Metrics & Citations  
Article History
Copyright
Rights and Permissions
Return

Comparison of prediction accuracies with DeepIso.

TaskDatasetAccuracy (%)
Charge stateNumber of spectrumsDeepIsoIsoFusion
Note: # represents that the predicted value has a certain tolerance; the best results are in bold.
Charge state110 64895.5699.92
254 67097.1899.61
333 08896.8199.63
436392.1998.93
526643.3699.42
All102 38096.6799.64
Number of isotopesAll102 38096.96#98.72#
Retention timeAll102 380−95.34#

Recall of high-confidence peptide detection. (%)

DatasetDinosaurMaxQuantPEAKSDeepIsoIsoFusion
Note: The best results are in bold.
130124_dilA_6_0195.8495.8196.3195.3896.69
130124_dilA_6_0295.5195.5296.6694.9697.32
130124_dilA_6_0395.1995.2696.0094.5097.11
130124_dilA_6_0495.3794.8695.6494.4897.70
130124_dilA_7_0194.8194.6195.4693.8196.89
130124_dilA_7_0295.7195.8096.5294.3997.75
130124_dilA_7_0395.5594.9096.3094.5797.68
130124_dilA_7_0495.6094.8996.1894.3997.52
Average95.4595.2196.1394.5697.33

Ablation study (“√” means the model has the corresponding setting).

NormFusionTrainingInferencePrediction accuracy
charge statem/zHardSoftHardSoftCharge stateNumber of isotopesRetention time
Note: Our IsoFusion model settings are highlighted in bold.
√–––√√–96.83% ± 0.0012%83.17% ± 0.0016%48.58% ± 0.0011%
√√––√√97.46% ± 0.0014%83.92% ± 0.0024%49.0% ± 0.0013%
√√√–√–√97.54% ± 0.0008%84.14% ± 0.0011%49.19% ± 0.0018%
√√√√–√–97.62% ± 0.0005%84.31% ± 0.0009%49.33% ± 0.0014%
–√√–√√–97.54% ± 0.0006%84.12% ± 0.0008%49.11% ± 0.0016%
√√√–√√–97.66% ± 0.0005%84.26% ± 0.0010%49.21% ± 0.0017%