输入Input
- 前体 m/zprecursor m/z
- 前体电荷precursor charge
- 碎片峰 m/zfragment peak m/z
- 碎片峰强度fragment peak intensity
- 原始谱图ID（仅用于关联预测）original spectrum ID (prediction alignment only)
比赛说明Competition Overview
把串联质谱图翻译成肽段序列Translate tandem mass spectra into peptide sequences
蛋白质是生命活动的重要执行者，其功能与氨基酸序列及结构密切相关。质谱能够高通量测量复杂生物样本中的肽段，但真正的挑战是从测量信号中读回序列。Proteins are major functional molecules in living systems, and their functions are closely related to amino acid sequence and structure. Mass spectrometry measures peptides at high throughput; the challenge is to recover sequence from the measured signal.
比赛目标Competition Goals
比赛的核心目标：开发能够准确地将 MS/MS 谱图转化为肽段序列的 AI 模型。Central goal: develop AI models that accurately translate MS/MS spectra into peptide sequences.
对于每一张未知谱图，模型需要预测产生它的完整 peptide sequence，并给出预测置信度。For every unknown spectrum, the model predicts the complete peptide sequence that generated it and reports a confidence score.
比赛任务Challenge Task
输入一张未知谱图，输出序列和置信度Input an unknown spectrum; output a sequence and confidence
这是一个 spectrum-to-sequence 生成任务：模型直接从谱图中的碎片峰信息恢复肽段序列；数据规模、文件与切分方式见 Dataset 页面。This is a spectrum-to-sequence generation task: directly recover a peptide sequence from fragment-peak information. Dataset scale, files and splits are described on the Dataset page.
AI 模型AI Model
架构不限。模型从谱图—肽段配对数据中学习碎裂规律、质量约束、强度分布与序列上下文。Any architecture. Learn fragmentation, mass constraints, intensity patterns and sequence context from spectrum–peptide pairs.
输出Output
- 预测肽段序列predicted peptide sequence
- 预测置信度prediction confidence score
数据库检索Database search
像选择题Like multiple choice
从预先定义的蛋白质数据库生成候选肽段，再选出与实验谱图最匹配的序列。Generate candidates from a predefined protein database, then select the sequence that best matches the spectrum.
VS
de novo 测序de novo sequencing
像开放题Like an open-ended question
不给候选答案，模型必须直接根据谱图生成完整的氨基酸序列。No candidate answers are provided; the model must generate the complete amino acid sequence directly from the spectrum.
比赛关注的研究方向Research directions
准确测序、泛化与高效推理Accurate sequencing, generalization and efficient inference
准确测序Accurate sequencing
能否从复杂、稀疏且带噪声的谱图中恢复完整氨基酸序列？Can AI reconstruct complete sequences from complex, sparse and noisy spectra?
真实泛化Generalization
面对未见肽段与独立实验项目，模型还能准确预测吗？Will performance hold on unseen peptides and independent experimental projects?
可复现性Reproducibility
冻结的代码与权重能否在统一环境中复现提交结果？Can frozen code and weights reproduce submitted predictions in a shared evaluation environment?
数据规模Data scaling
训练对从百万级增长到千万、亿级时，性能能否持续提升？How does performance change from millions to hundreds of millions of training pairs?
高效 AIEfficient AI
能否兼顾准确率、模型规模、推理速度与计算成本？Can accuracy be balanced with model size, inference time and compute?
为什么需要 de novo sequencing？Why de novo sequencing?
从“在数据库里找答案”走向“发现未知序列”From searching a database to discovering unknown sequences
数据库搜索只能在预先定义的蛋白质空间中匹配候选肽段；真实肽段若未被收录，答案就不可能出现。de novo 直接从谱图生成序列，为新肽段、序列变异、非常规翻译产物、免疫肽和注释不足物种打开探索空间。这些是潜在应用方向，并非本届比赛的独立赛道。Database search can only match candidates in a predefined protein space. If a real peptide is absent, it cannot be returned. de novo generation opens discovery space for novel peptides, variants, non-canonical products, immunopeptides and poorly annotated organisms. These are potential applications, not separate competition tracks.
面向 AI 研究者An AI perspective
这是一个带有物理约束的 spectrum-to-sequence 生成问题A spectrum-to-sequence generation problem with physical constraints
从 AI 的角度，蛋白质组学中的概念可以直接映射到熟悉的信号、序列和域泛化问题。From an AI perspective, proteomics maps directly onto familiar problems in signals, sequences and domain generalization.
蛋白质组学Proteomics
AI 视角AI Perspective
MS/MS 谱图MS/MS spectrum
稀疏连续信号Sparse continuous signal
谱图峰Spectrum peak
观测 / 标记（token）Observation / token
m/z
连续位置Continuous position
峰强度Peak intensity
信号强度Signal strength
肽段Peptide
目标序列Target sequence
氨基酸Amino acid
词表Vocabulary
前体质量Precursor mass
全局物理约束Global physical constraint
碎裂规则Fragmentation rule
物理归纳偏置Physical inductive bias
仪器Instrument
技术域Technical domain
物种Species
生物域Biological domain
de novo 测序de novo sequencing
谱图到序列生成Spectrum-to-sequence generation
因此，无论你的研究背景是 Transformer、LLM、sequence modeling、computer vision、diffusion model、representation learning、graph neural network 或 foundation model，都可以参与这一问题。Whether your background is in Transformers, LLMs, sequence modeling, computer vision, diffusion models, representation learning, graph neural networks or foundation models, you can contribute to this problem.
为什么这是一个有挑战性的 AI 问题？Why is this a challenging AI problem?
稀疏、连续且带噪声的输入Sparse, continuous and noisy input
变长序列生成与结构化预测Variable-length structured generation
质量守恒等物理约束Physical mass constraints
仪器、物种与项目域偏移Instrument, species and project shifts
分布外泛化与序列防泄漏OOD generalization and leakage control
固定模型的可复现推理Reproducible inference with frozen models
accuracy × generalization × reproducibility × efficiency
de novo 肽段测序de novo peptide sequencing
四步理解肽段从头测序Understand de novo sequencing in four steps
你不需要先学习完整的蛋白质组学课程。先理解蛋白质如何变成一张谱图，以及 AI 如何反向读回序列。You do not need a full proteomics course. Start with how a peptide becomes a spectrum, then see how AI solves the process in reverse.
蛋白质被切成肽段Protein becomes peptides
蛋白质由氨基酸组成，实验中通常先被切成较短的 peptide。例如 PEPTIDE 由 7 个氨基酸残基组成。Proteins consist of amino acids and are commonly broken into shorter peptides. PEPTIDE has seven residues.
质谱仪使肽段碎裂The peptide is fragmented
选中的 peptide ion 在碰撞下断裂，形成具有规律性的 fragment ions，包括常见的 b ions 和 y ions。A selected peptide ion is fragmented, producing characteristic fragment-ion series such as b and y ions.
碎片形成 MS/MS 谱图Fragments form an MS/MS spectrum
仪器测量碎片的 m/z 和 intensity。氨基酸质量不同，因此相关峰之间的质量差携带序列信息。The instrument measures fragment m/z and intensity. Mass differences between related peaks carry sequence information.
AI 反向生成序列AI generates the sequence
模型从大量 spectrum–peptide pairs 学习碎裂和质量规律，再从未知谱图直接生成 amino acid sequence。The model learns fragmentation and mass relationships from spectrum–peptide pairs, then generates a sequence for an unseen spectrum.
这是一个逆问题This is an inverse problem
正向实验流程Forward experiment
肽段序列Peptide sequence→碎裂与质谱测量Fragmentation + MS measurement→MS/MS 谱图MS/MS spectrum
AI 逆向推断任务AI inference task
MS/MS 谱图MS/MS spectrum→AI 模型AI model→肽段序列 + 置信度Peptide sequence + confidence
继续学习Keep learning
从基础概念到最新模型From fundamentals to current models
下面的公开资料适合没有蛋白质组学背景的参赛者按顺序阅读。These public resources are sequenced for participants without a proteomics background.
先看基础Start with the basics
EMBL-EBI · Introduction to Proteomics
EMBL-EBI · MS-based Proteomics
阅读代表性模型Read representative models
DeepNovo · PubMed
Casanovo · ICML 2022
NovoBench · benchmark survey
动手实践Try it yourself
Casanovo · code & documentation
InstaNovo · code & model
再参考本赛题的基线、评测说明与提交示例。Then consult the competition baseline, evaluation guide and submission example.
1亿条训练PSM，统一的数据划分100 million training PSMs, consistent evaluation splits
训练集Training
100M1亿条PSM；417个Parquet文件。来自527个项目、14,208个源文件，覆盖38个标准化物种/分类标签和10类仪器标签。100 million PSMs in 417 Parquet files, from 527 projects and 14,208 source files, covering 38 standardized species/taxonomic labels and 10 instrument labels.
公开测试集Public test
3M300万条PSM。提供谱图和参考答案，用于开发反馈及本地评估。物种均在训练数据中出现，但项目完全隔离。3 million PSMs. Spectra and reference labels support development and local evaluation. Species are represented in training, but projects are disjoint.
隐藏测试集Hidden test
模型冻结并完成材料核验后开放谱图。参考答案不向选手开放；具体物种和项目构成不予公开。Spectra are released after model freezing and package verification. Reference labels remain private; species and project composition are not disclosed.
数据隔离与泛化Isolation and generalization
训练、公开测试与隐藏测试在项目、源文件和标准化肽段序列（I/L视为等价、不含修饰标注）三个层面两两完全隔离，同一项目、源文件或标准化肽段序列不会跨集合出现。两个测试集的仪器条件均包含在训练数据覆盖范围内，不设置未见仪器条件。隐藏测试用于检验独立实验项目和未见肽段上的泛化能力。Training, public test and hidden test are pairwise disjoint by project, source file and normalized peptide sequence (I/L equivalent, modification annotations removed). Both test sets use instrument conditions represented in training; there is no unseen-instrument condition. Hidden evaluation tests generalization to independent projects and unseen peptides.
文件与下载Files and downloads
数据以Parquet格式通过Google Drive分发，支持分批读取。模型输入包括碎片峰m/z、峰强度、前体m/z和前体电荷。谱图原始spectrum_id用于关联预测，不得用于检索答案。训练及Public数据在开赛时开放；隐藏谱图按赛程另行发布。Parquet data are distributed through Google Drive and support batch reading. Inputs include fragment m/z and intensity, precursor m/z and charge. Original spectrum_id values align predictions and must not be used to retrieve answers. Training and public data open at launch; hidden spectra follow the staged timeline.
数据说明与Google Drive下载入口Data guide and Google Drive downloads以完整肽段预测衡量模型表现Evaluate complete peptide predictions
主评分为按物种等权平均的肽段召回率，分数越高越好。公开榜用于开发反馈，隐藏测试的Private成绩用于最终排名，并须通过复现与合规审查。The primary metric is species-macro peptide recall; higher is better. Public scores provide development feedback. Hidden-test Private scores determine final ranking, subject to reproduction and compliance review.
主评分：物种宏平均肽段召回率Primary metric: species-macro peptide recall
先计算每个物种完整肽段预测正确的谱图记录数占该物种参考记录总数的比例，再对所有物种取算术平均。每条谱图分别计分，不按唯一肽段去重；NO_PREDICTION计为错误且保留在分母中。For each species, divide the number of spectrum records with a correctly predicted complete peptide by all reference records for that species, then take the arithmetic mean across species. Every spectrum counts; peptide sequences are not deduplicated. NO_PREDICTION counts as incorrect and remains in the denominator.
完整骨架匹配Complete backbone matching
预测和参考序列按相同规则归一化，完整肽段骨架精确一致才算正确。I/L视为等价；支持的数值型及UNIMOD修饰注释在比较前移除，不对修饰类型或位点评分。不支持的注释格式将报错。Predicted and reference sequences undergo identical normalization and must match exactly across the complete backbone. I/L are equivalent. Supported numeric and UNIMOD modification annotations are removed; modification type and position are not scored. Unsupported notation is rejected.
置信度与运行效率Confidence and runtime
提交需包含有限数值的置信度，但主评分不按置信度加权，也不设置预测置信度或1% FDR截断。推理时间是获奖资格要求，不加权计入主评分，也不用于近似并列时的排名。具体运行条件及上限详见Kaggle比赛页面。A finite numeric confidence score is required, but it does not weight the primary metric or impose a confidence or 1% FDR cutoff. Runtime is an award-eligibility requirement, not a score weight or a near-tie ranking criterion. See Kaggle for the runtime limit and testing conditions.
查看完整评测说明Full evaluation details参赛规则Competition rules
参赛资格Eligibility
允许个人或团队参赛，每队最多5人，面向学生、科研人员、工程师及产业团队。参与者仍须遵守平台适用的参与条款。Individuals and teams of up to five may enter, including students, researchers, engineers and industry teams, subject to applicable platform participation terms.
外部数据与模型External data and models
允许外部训练数据和外部预训练模型，须披露来源、版本、许可及已知数据重叠。模型架构不限；不另设限制外部数据的主赛道或研究赛道。External training data and pretrained models are allowed with disclosure of sources, versions, licenses and known data overlaps. Any model architecture is allowed; there is no separate restricted-data or research track.
从头测序与防作弊De novo inference and integrity
推理阶段不得通过蛋白质/肽段数据库、谱图库或数据库搜索生成、检索、筛选或重排候选答案。禁止将隐藏测试用于训练、人工标注测试谱图、利用测试集身份或推断出的物种设置答案检索组件、交换私有预测，或用多账号绕过提交限制。违规取得的成绩不予认可。Inference must not use protein/peptide databases, spectral libraries or database search to generate, retrieve, filter or rerank candidate answers. Training on hidden test data, manual test annotation, answer-retrieval components keyed to test identity or inferred species, exchanging private predictions and bypassing submission limits through multiple accounts are prohibited. Scores obtained in violation of the rules will not be recognized.
模型冻结与复现Model freezing and reproduction
参与最终评测的队伍须在隐藏谱图开放前提交并冻结完整代码、模型权重、配置和运行环境。核验通过后只能用冻结的推理流程生成最终预测。主办方复现预测并审查训练数据来源；无法复现或违反规则的结果不具备最终获奖资格。Teams entering final evaluation must submit and freeze complete code, weights, configuration and runtime environment before hidden spectra are released. Approved teams generate final predictions only with the frozen pipeline. Organizers reproduce predictions and audit training-data sources; non-reproducible or non-compliant results are ineligible for awards.
提交格式与排行榜Submission format and leaderboards
Public和最终评测使用同一个Kaggle比赛。Public阶段CSV保留原始spectrum_id，覆盖全部300万条Public记录，无需补齐隐藏阶段数据。列名如下；无法预测时序列填NO_PREDICTION，置信度填0。隐藏阶段发布后提交该阶段的完整预测。每日提交次数等平台限制以比赛页面为准。Public and final evaluation use the same Kaggle competition. Public CSVs retain original spectrum_id values and cover all 3 million public records, without padding hidden rows. Use the columns below; enter NO_PREDICTION and confidence 0 for abstentions. After hidden release, submit complete predictions for that stage. Daily submission limits are specified on Kaggle.
spectrum_idpredicted_peptide_sequenceconfidence_score
格式说明与Public示例Format and public submission example17万元奖池，6个月挑战CNY 170,000 in prizes, a six-month challenge
报名及比赛于2026年9月26日10:00开放，至2027年3月26日24:00（即3月27日00:00）截止，正式赛期6个月。所有时间均为北京时间（UTC+8）。Registration and competition open on 26 September 2026 at 10:00 and close on 27 March 2027 at 00:00 (the end of 26 March). The competition lasts six months. All times are Beijing time (UTC+8).
总奖池人民币170,000元，共奖励6支参赛队伍（含个人参赛者）。奖金均为税前金额，以人民币发放。Total prize pool: CNY 170,000 for six teams (including individual entrants). All amounts are before tax and paid in CNY.
二等奖Second prize¥30,0002队 · 每队金额2 team(s) · amount per team
一等奖First prize¥50,0001队 · 每队金额1 team(s) · amount per team
三等奖Third prize¥20,0003队 · 每队金额3 team(s) · amount per team
最终获奖结果须通过模型复现与合规审查。赛事也希望创造实习推荐、科研交流与合作机会；如有相关安排，将另行通知。Awards are subject to reproduction and compliance review. The competition also seeks to facilitate internship referrals, research exchange and collaboration; any opportunities will be announced separately.
谁可以参加？Who can participate?
欢迎人工智能、机器学习、数据科学、生物信息学、蛋白质组学等相关领域的学生、科研人员、工程师及产业团队。允许个人或最多5人团队参赛，须遵守平台适用条款。Students, researchers, engineers and industry teams in AI, machine learning, data science, bioinformatics, proteomics and related fields are welcome. Enter individually or as a team of up to five, subject to platform terms.
前往Kaggle · 9月26日10:00开放报名Kaggle · registration opens 26 September, 10:00基线与开发支持Baseline and development support
XuanjiNovo 100M预训练模型在全部300万条公开测试谱图上的物种宏平均肽段骨架召回率为58.03%。代码及预训练模型已开源，可在允许的预训练模型基础上改进，也可开发其他模型。比赛页面提供评测说明及CSV提交示例。The XuanjiNovo 100M pretrained baseline achieves 58.03% species-macro peptide backbone recall on all 3 million public spectra. Code and pretrained weights are open source. Teams may improve permitted pretrained models or develop other models. Evaluation documentation and CSV examples are provided on Kaggle.
XuanjiNovo开源仓库XuanjiNovo repository评测与提交说明Evaluation and submission guide赛事时间线 · 北京时间 UTC+8Timeline · Beijing time UTC+8
报名与开发开始Registration and development
2026-09-26 10:00
训练及Public数据通过Google Drive开放。Training and public data open via Google Drive.
模型收集Model collection
≈ 2027-02-10
开发与收集并行，截止前允许更新。Development continues; updates are allowed until freezing.
模型冻结Model freeze
≈ 2027-03-10
统一截止并冻结，预留约2天核验最终版本。Final packages freeze; approximately two days for verification.
隐藏谱图开放Hidden spectra release
2027-03-12
合格队伍使用冻结流程，最后约半个月进行推理。Approved teams infer with frozen pipelines for approximately two weeks.
最终预测截止Final prediction deadline
2027-03-26 24:00
即3月27日00:00；赛后开展复现审核。Equivalent to 27 March 00:00; reproduction follows.
赛后复现审核按提交量和实际工作量安排；审核及适用复核完成后公示结果、颁奖，具体时间另行通知。如遇特殊情况调整赛程，以最终比赛公告为准。Post-competition reproduction depends on submission volume and workload. Results and awards follow review and any applicable appeals; dates will be announced separately. Schedule changes will be communicated in official competition announcements.
报名与咨询Registration and questions
报名于2026年9月26日10:00（北京时间）开放。Public开发与最终评测均在同一个Kaggle页面进行。联系邮箱：ajun@westlake.edu.cn · zhangwenjie@westlake.edu.cn。赛事更新请关注比赛页面公告与讨论区。Registration opens on 26 September 2026 at 10:00 Beijing time. Public development and final evaluation share one Kaggle page. Contact: ajun@westlake.edu.cn · zhangwenjie@westlake.edu.cn. See competition announcements and discussions for updates.
比赛公告与讨论区Announcements and discussions组织单位Organizers
西湖大学Westlake University
布鲁克Bruker
常见问题Frequently asked questions
什么时候报名？When can I register?
2026年9月26日10:00（北京时间）在Kaggle开放报名。最终预测于2027年3月26日24:00截止。Registration opens on Kaggle on 26 September 2026 at 10:00 Beijing time. Final predictions are due at the end of 26 March 2027.
能用外部数据和预训练模型吗？Are external data and pretrained models allowed?
允许，须披露来源、版本、许可及已知数据重叠。不得利用隐藏测试训练或通过答案检索完成推理。Yes, with source, version, license and known-overlap disclosure. Hidden-test training and answer-retrieval inference are prohibited.
主办方提供GPU吗？Will GPUs be provided?
目前不承诺参赛算力支持；如有安排将另行通知。选手可使用自行准备的计算资源。Participant compute support is not currently committed. Any arrangements will be announced separately; teams may use their own resources.
置信度影响排名吗？Does confidence affect ranking?
当前主评分不按置信度加权，也不设1% FDR截断。仍须按格式提交有限数值的置信度。The primary score uses neither confidence weighting nor a 1% FDR cutoff. A finite numeric confidence field remains required.
开赛就能获得隐藏数据吗？Is hidden data available at launch?
不能。先收集、冻结并核验模型包，再按时间线开放隐藏谱图。隐藏答案不向选手提供，最终结果须经复现审查。No. Model packages are collected, frozen and verified before hidden spectra are released. Hidden labels remain private, and final results undergo reproduction review.
Public提交需要补齐隐藏记录吗？Do public submissions need hidden rows?
不需要。提交全部300万条Public谱图对应的预测即可，保留原始ID；不包含其他阶段记录。No. Submit predictions for all 3 million public spectra with original IDs, without rows from other stages.