jingbo02 commited on
Commit
9c7f4db
·
verified ·
1 Parent(s): fb36f99
Files changed (1) hide show
  1. README.md +70 -3
README.md CHANGED
@@ -1,3 +1,70 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ ---
4
+
5
+ # Dataset Card for NovoBench
6
+ Datasets used for the baseline comparison of deep learning-based de novo peptide sequencing method
7
+
8
+ ## Dataset Description
9
+
10
+ - **Repository:** [NovoBench](https://github.com/jingbo02/NovoBench)
11
+ - **Paper:** [NovoBench: Benchmarking Deep Learning-based De Novo Peptide Sequencing Methods in Proteomics](https://arxiv.org/pdf/2406.11906)
12
+
13
+
14
+ ### Dataset Summary
15
+
16
+ - **Seven-species Dataset.** The Seven-species dataset contains low-resolution mass spectrum and
17
+ their peptide labels from 7 different species. The previous work DeepNovo has evaluated its
18
+ performance on these datasets with the leave-one-out method, i.e., training the model on 6 species and
19
+ testing on the left one species, to mimic the real-world challenging cases where we have to identify
20
+ the never-before-seen peptide sequences for the observed mass spectrum. In this paper, we conducted
21
+ testing on the yeast species and training on the remaining 6 species.
22
+
23
+ - **Nine-species Dataset.** The Nine-species dataset is the most widely-used dataset by previous works
24
+ such as DeepNovo, PointNovo, and Casanovo, which contains high-resolution mass
25
+ spectrum and their peptide labels from 9 different species. We adopt the Nine-species dataset
26
+ used by the original publication of DeepNovo (MassIVE dataset identifier: MSV000081382) for
27
+ benchmarking. Similar to Seven-species dataset, we train models on 8 species and evaluate the left
28
+ yeast species. Additionally, these datasets contain 3 PTMs (oxidation of methionine, deamidation of
29
+ asparagine or glutamine), enabling the fair evaluation of various models’ performance in terms of
30
+ identifying PTMs.
31
+ - **HC-PT Dataset.** The HC-PT dataset, as detailed in the InstaNovo paper, includes synthetic tryptic
32
+ peptides that span all canonical human proteins and isoforms. It also encompasses peptides generated
33
+ by alternative proteases and HLA peptides. The key feature of the HC-PT dataset is its high-resolution
34
+ spectrum for human-origin peptides and the peptide labels are derived from the high-confidence
35
+ search results of MaxQuant.
36
+
37
+
38
+ ## Dataset Structure
39
+
40
+ The dataset is tabular, where each row corresponds to a labelled MS2 spectra.
41
+ - `sequence (string)` \
42
+ The target peptide sequence excluding post-translational modifications
43
+ - `modified_sequence (string)` \
44
+ The target peptide sequence including post-translational modifications
45
+ - `precursor_mz (float64)` \
46
+ The mass-to-charge of the precursor (from MS1)
47
+ - `charge (int64)` \
48
+ The charge of the precursor (from MS1)
49
+ - `mz_array (list[float64])` \
50
+ The mass-to-charge values of the MS2 spectrum
51
+ - `mz_array (list[float32])` \
52
+ The intensity values of the MS2 spectrum
53
+
54
+
55
+
56
+ ## Citation Information
57
+
58
+ If you use this dataset, please cite the NovoBench:
59
+
60
+ ```bibtex
61
+ @misc{zhou2024novobenchbenchmarkingdeeplearningbased,
62
+ title={NovoBench: Benchmarking Deep Learning-based De Novo Peptide Sequencing Methods in Proteomics},
63
+ author={Jingbo Zhou and Shaorong Chen and Jun Xia and Sizhe Liu and Tianze Ling and Wenjie Du and Yue Liu and Jianwei Yin and Stan Z. Li},
64
+ year={2024},
65
+ eprint={2406.11906},
66
+ archivePrefix={arXiv},
67
+ primaryClass={q-bio.QM},
68
+ url={https://arxiv.org/abs/2406.11906},
69
+ }
70
+ ```