Foundation models come to mass spectrometry proteomics Identifying proteins from their fragments is a foundational task in biology. A mass spectrometer breaks peptides into pieces, measures the masses, and software then tries to match each spectrum back to a peptide sequence. Two approaches have dominated for decades: database search (match against known proteomes) and de novo sequencing (infer the peptide directly from the spectrum). Both are bottlenecked by the same step, scoring how well a candidate peptide explains a spectrum. Deep learning has been entering this field for years, mostly as feature extractors feeding traditional engines like MaxQuant or MSFragger. Jiale Zhao and coauthors introduce pUniFind, is a multimodal foundation model trained on over 100 million peptide-spectrum matches from open database searches. Spectra and peptides get their own encoders, and the model is pretrained with cross-modality tasks: predict the spectrum from the peptide, predict the peptide from the spectrum, score peptide-spectrum pairs jointly. Database search and de novo sequencing become two views of the same model. The numbers are remarkable. In immunopeptidomics, where peptides come from non-tryptic digestion and are very hard to identify, pUniFind finds 42.6% more peptides than Open-pFind. In modification-rich de novo sequencing, it identifies 60% more peptide-spectrum matches than existing methods, despite working in a 300 times larger search space. In regular de novo, it recovers 38.5% more peptides, including 1,891 that map to the human genome but are absent from reference proteomes. A deep learning quality control filter raises consistency with RNA-Seq evidence from 65.4% to 85.0%. What makes this an ML story is the architecture choice. End-to-end scoring with a shared latent space replaces hand-crafted feature pipelines, and the same backbone serves both database search and de novo sequencing without retraining. For drug discovery, immunotherapy, and antibody engineering, peptide identification is the bottleneck for neoantigen discovery and biomarker pipelines. Tools that find more peptides at the same FDR, especially in non-tryptic and modification-rich settings, expand what is detectable from existing data. That changes what is worth running in the lab. Paper: Zhao et al., Nature Machine Intelligence (2026) — journal license | doi.org/10.1038/s42256-026-0…

Six papers worth your Sunday coffee. ⚽ 𝗧𝗵𝗲 𝗪𝗼𝗿𝗹𝗱 𝗖𝗵𝗮𝗺𝗽𝗶𝗼𝗻'𝘀 𝗣𝗼𝗿𝘁𝗳𝗼𝗹𝗶𝗼: 𝗔 𝗙𝗮𝗰𝘁𝗼𝗿-𝗕𝗮𝘀𝗲𝗱 𝗜𝗻𝘁𝗲𝗿𝗽𝗿𝗲𝘁𝗮𝘁𝗶𝗼𝗻 𝗼𝗳 𝗠𝗼𝗱𝗲𝗿𝗻 𝗙𝗼𝗼𝘁𝗯𝗮𝗹𝗹 University of Zurich, Swisscanto A starting eleven read as a factor portfolio. Defence is Value, midfield is Quality, attack is Momentum, the playmaker is the adaptive AI model, the keeper is the risk manager. Satirical, but the analogy holds up better than it has any right to. 🕸️ 𝗚𝗲𝗼𝗺𝗛𝗲𝗿𝗱: 𝗙𝗼𝗿𝘄𝗮𝗿𝗱-𝗹𝗼𝗼𝗸𝗶𝗻𝗴 𝗛𝗲𝗿𝗱𝗶𝗻𝗴 𝘃𝗶𝗮 𝗥𝗶𝗰𝗰𝗶 𝗙𝗹𝗼𝘄 𝗼𝗻 𝗔𝗴𝗲𝗻𝘁 𝗚𝗿𝗮𝗽𝗵𝘀 Imperial College London, USTC, MaxQuant, HKU, UCL Ollivier-Ricci curvature on LLM agent interaction graphs catches herding before it hits realised returns. Median lead of 272 simulator steps before the order parameter crosses. 🤖 𝗔𝗴𝗲𝗻𝘁𝘀 𝗔𝗿𝗲 𝗡𝗼𝘁 𝗔𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺𝘀: 𝗧𝗿𝗮𝗱𝗲𝗼𝗳𝗳𝘀 𝗼𝗳 𝗗𝗲𝗰𝗶𝘀𝗶𝗼𝗻-𝗧𝗶𝗺𝗲 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 𝗶𝗻 𝗔𝗜 𝗧𝗿𝗮𝗱𝗶𝗻𝗴 Rotman, RBC Capital Markets Frontier LLMs running tender selection and execution on the Rotman simulator. More reasoning earns a dividend on selection but pays a tax in expired intents and stuck inventory. Market speed decides which side wins. 📈 𝗬𝗶𝗲𝗹𝗱 𝗖𝘂𝗿𝘃𝗲 𝗗𝘆𝗻𝗮𝗺𝗶𝗰𝘀 𝗨𝘀𝗶𝗻𝗴 𝗩𝗔𝗘𝘀 𝗨𝗻𝗱𝗲𝗿 𝗡𝗼-𝗔𝗿𝗯𝗶𝘁𝗿𝗮𝗴𝗲 Johns Hopkins A Student-t Conditional VAE learns a heavy-tailed yield curve manifold. A Neural SDE evolves it under a PDE no-arbitrage penalty. Out-of-sample mean tenor RMSE of 6.58 bps across USD, GBP and JPY, zero-lower-bound respected. 🧱 𝗘𝗻𝗵𝗮𝗻𝗰𝗶𝗻𝗴 𝗮 𝗥𝗶𝘀𝗸 𝗠𝗼𝗱𝗲𝗹 𝗯𝘆 𝗔𝗱𝗱𝗶𝗻𝗴 𝗧𝗿𝗮𝗻𝘀𝗶𝗲𝗻𝘁 𝗦𝘁𝗮𝘁𝗶𝘀𝘁𝗶𝗰𝗮𝗹 𝗙𝗮𝗰𝘁𝗼𝗿𝘀 Stanford, BlackRock (Candès, Hastie, Boyd, Kochenderfer, Kahn) A weighted maximum-likelihood EM procedure that refines the Barra short-term US risk model and bolts on a handful of statistical factors. Picks up regime shifts a quarterly update would miss. 🎲 𝗖𝗿𝘆𝗽𝘁𝗼𝗰𝘂𝗿𝗿𝗲𝗻𝗰𝘆 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻 𝗠𝗮𝗿𝗸𝗲𝘁𝘀 𝘁𝗵𝗿𝗼𝘂𝗴𝗵 𝘁𝗵𝗲 𝗗𝗲𝗿𝗶𝘃𝗮𝘁𝗶𝘃𝗲𝘀 𝗟𝗲𝗻𝘀 Seoul National University Treats Kalshi and Polymarket binaries as cash-or-nothing digitals and extracts the full toolkit: implied vol surface, risk-neutral density, variance risk premium. Polymarket's VRP is roughly 16x Kalshi's, mostly contract design rather than mispricing. I started a monthly digest that cuts out all the noise in #Quant_Finance & #AI/#LLM research. Just the things that actually matter. Friends keep telling me it saves them a ton of time. Sign up here: forms.gle/a7KpFxZN4xiz3ZaMA Here are past issues if you want a peek: zanista.ai/paperpal

我把23年-26年这4届香港web3嘉年华赞助商名单喂给了ai，让他分析下行业趋势和热点，感觉蛮有意思的： ———————————— 从 2023 年到 2026 年这四年的香港 Web3 嘉年华赞助商名单，不仅是企业的变迁史，更是 Web3 行业从“废墟重建”到“生态大爆发”，再到“专业化与 AI 融合”的缩影。 通过这份名单，我们可以把行业的演进逻辑拆解为以下四个阶段： 2023年：筑基与信任重建（关键词：基建、安全、合规） 行业背景： 刚经历过 2022 年的 FTX 和 Luna 暴雷，全行业处于“去杠杆”后的修复期。 赞助商特征： 安全公司霸榜： Beosin、慢雾、Safeheron 悉数在列。大家当时最关心的是“活下去”和“资产安全”。 老牌基建稳场： OKX、Neo、Avalanche、Nervos。这些是行业的“老钱”和底层支柱，在深熊中展现了韧性。 大厂试水： 腾讯云（Tencent Cloud）的出现，标志着传统云服务商开始关注 Web3 基础设施。 明星项目： OKX（确立了其在香港市场的绝对领导地位）、Avalanche、Nervos。 2024年：生态爆发与应用元年（关键词：TON、比特币生态、RWA/合规金服） 行业背景： 比特币现货 ETF 通过，牛市预期回归，生态开始多元化。 赞助商特征： TON 的强势崛起： TON Foundation 和相关生态项目开始占据核心位置。Telegram 的巨大流量入口成为全行业焦点。 比特币生态回暖： Nervos CKB 转变为 BTC L2 逻辑，以及 Bitcoin Magazine 的参与，反映了 BTC 生态的热度。 合规力量入场： 众安银行（ZA Bank）、胜利证券的出现，说明香港的监管框架吸引了正规军。 云服务商全面接管： AWS、Google Cloud、阿里云、腾讯云“四大金刚”齐聚。Web3 已经不再是小作坊，而是进入了大厂云服务的时代。 明星项目： TON、Chainlink、Solana 生态项目。 2025年：流量为王与交易工具化（关键词：GMGN、土狗/Meme、极简体验） 行业背景： 散户情绪高涨，Meme 币和链上交易成为主流。 赞助商特征： 流量与工具的胜利： GMGN 成为冠名赞助商。这标志着行业从“高大上的协议”转向“实用的链上交易工具”。 专注“搞钱”： Alchemy Pay、OneKey（硬件钱包）的持续稳健，说明支付和出入金依然是硬需求。 生态收割： TON 依然稳坐冠名位置，生态红利持续释放。 明星项目： TON、GMGN、SlowMist（慢雾已成为行业长青标杆）。 2026年：专业化、AI 与跨界融合（关键词：AI 驱动、专业交易、机器人） 行业背景： 行业进入深水区，AI 与 Web3 的融合从口号变为产品，专业投资者比例上升。 赞助商特征： AI 浓度极高： Finanx AI、ChainSafeAI、BytePlus（字节跳动旗下）的出现，说明行业正在利用 AI 进行量化、风控和内容生成。 专业化交易工具： SignalPlus 跻身标题赞助商，反映了期权、波动率交易等专业金融工具在香港这个金融中心的需求激增。 跨界与硬件： HK ROBOTICS（香港机器人）和字节跳动的 BytePlus 参与，说明 Web3 正在与实体科技、大数据边缘计算融合。 细分领域的深度： 出现了针对特定赛道的赞助，如“以太坊应用协会”、“Web3 Leader Program”。 明星项目： SignalPlus（专业交易员首选）、Sui（晚宴赞助，展示财力与生态自信）、TRON。 总结：四年演进的底层逻辑 从“链”到“应用”再到“工具”： 23年大家在聊哪条公链牛，24年在聊哪个生态大，26年已经在聊哪款 AI 工具能帮我跑赢市场。 Web2 巨头的角色转变： 23-24年：云服务提供商（卖水人）。 25-26年：技术架构深度参与者（BytePlus、Google Cloud 的高频出现）。 专业化的必然性： 23年的赞助商多是“通用型”。 26年的赞助商名单里，SignalPlus（期权交易）、MaxQuant（量化）、Finanx AI（金融AI） 的位置前移。这意味着市场参与者已经从单纯的“持币者”演变为“专业化交易员/策略师”。 一句话总结：这四年的变化，是香港 Web3 从“寻求认可的拓荒者”，变为了“专业金融与前沿 AI 结合的试验场”。 @SignalPlusCN

🐙On‑Site Challenge🐙 📸 “Snapped a pic at MaxQuant?” 🎁 “Claimed your trial bonus?” 🎮 “Tried Octopus?” Now turn it into a tweet and join the challenge! 📅 April 21 – 23 ✅ How to enter: Quote / RT this post Post your MaxQuant photo (booth, swag, anything MaxQuant-related) Include your UID tag @DeepTradeX_ & 1 friend 🔁 Your tweet = your entry. 🎁 Rewards: 10 winners × 5 USDT each. Winners will be randomly selected after April 23 and sent to your DeepTradeX account within 7 days. 👉 Grab your trial bonus: deeptradex.ai #Web3 #AITrading #Crypto #QuantTrading #HongKong #DeepTradeX #MaxQuant #AIvsHuman

🎮 “How do I claim the trial bonus?” 🎮 “How do I use Octopus?” 🎮 “What’s the latest?” ✨ You asked, we answered. 📈 Live from the floor: The venue was packed right after doors opened. The MaxQuant zone kept getting busier by the minute. ✅ Zero‑risk trial is live. Grab your bonus and full guide now: deeptradex.ai/ 👉 The line is only getting longer. Come say hi at our booth and join the fun! #Web3 #AITrading #Crypto #QuantTrading #HongKong #DeepTradeX #MaxQuant #AIvsHuman

AI Trading is no longer a concept — it’s happening live in Hong Kong. ⚡ Join @MaxQuant_AI × @DeepTradeX_ at #Web3Festival and experience: 📱 AI trading in 30 seconds 🧠 Transparent decision-making ⚔️ AI vs Human in real market conditions 🗓 April 21st, 10:30–12:30 (GMT 8) 📍 Grand Hyatt Hong Kong Scan the QR code and secure your spot now. This isn’t a demo. This is AI trading — live. #Web3 #AITrading #Crypto #QuantTrading #HongKong #DeepTradeX #MaxQuant #AIvsHuman

🎉We Reached $130M trading volume in March. 300% capital utilization.With almost zero marketing. While most are still building research-focused agents, our Agentic Trader executes end-to-end: strategy → execution → risk control. Always in the market. Always trading smarter. #AITrading #AgenticTrader #AIAgent #Web3 #Crypto #MaxQuant

@Olizzi9 @lana17500 @olivia_recchi @MonaLisas45 @asumcaz @uHrQqODpCxiU9IC 😎😎

We're proud to welcome @MaxQuant_AI as a Gold Sponsor of Hong Kong Web3 Festival 2026. Built on the AgentFi protocol, MaxQuant AI is a new kind of financial infrastructure, an intelligent trading partner. More information: maxquant.ai/ 🗓️20-23, April | HKCEC - 5BCDE 🎫Tickets: luma.com/hkweb3festival_2026 #Web3Festival #AI

AI trading = algorithmic trading machine intelligence. Real-time data processing. Pattern recognition. Adaptive strategies. Automated execution. With machine learning, trading systems don’t just follow rules — they evolve with the market. Discover AI trading with MaxQuant AI today, at maxquant.ai #aitrading #tradingagent

QA Bot Update → Agent 1.1 Preview After QA Agent 1.0 infra is ready, we’re moving beyond pure trade actions. Agent 1.1 = Graph-first Summary → text highlights 2–3 composite charts → “1 min to understand what happened & what’s next” Core outputs: • Price Canvas: price events importance KOL sentiment ( on/off-chain) • Liquidity Canvas: liquidation heatmap same event markers Workflow: multi-source signals → time-align → score → de-conflict → rank → stable visual summary, every time. Trade smarter with MaxQuant AI. New version will be available soon.

European markets are finally stabilizing after days of tech-driven sell-offs. All eyes now turn to Nvidia’s earnings to gauge whether the AI trade still has fuel. Meanwhile, UK inflation drops to 3.6%, boosting rate-cut hopes. A critical day for global sentiment. MaxQuant AI is signaling a BUY on NVDA. #Stocks #Nvidia #AI #EuropeMarkets #Inflation #BoE #Finance #Investing #MaxQuantAI

The world’s largest company, Nvidia is set to release its Q3 results today. MaxQuant AI is signaling a #BUY on $NVDA. #NVDA #Nvidia #AIAgent #StockMarket #StockMarketUpdate #StocksToBuy #MaxQuantAI

BTC long opened at 110,129 and closed at 110,228. 40× leverage | 9-minute hold. Quick precision by MaxQuant AI. ⚙️ #Bitcoin #CryptoMarket #AITrading #Crypto

BTC long opened at 110,330. Market pulled back, but MaxQuant AI gave another buy signal at 109,964. Both positions closed at 110,496. 40× leverage | 12h 14min hold. #Bitcoin #Crypto #CryptoMarket #AITrading #Trading

MaxQuant AI - built on the AgentFi protocol, a new kind of financial infrastructure — your intelligent trading partner. 📊 Analyze US stocks & crypto markets 🧠 Plan. Execute. Optimize. Welcome to smarter investing. #CryptoMarket #StockMarket #AI #Trading

a SELL signal been detected by MaxQuant AI. Entry price：109,518 20x short Exit price:109,290 Holding time: 14 mins Can we join the game? @jay_azhang @the_nof1 #Bitcoin #crypto #MaxQuantAI #AITrading #Trading

Another day another win. MaxQuant AI definitely smashed the game! Entry Price:186.97 Exit Price: 187.46 20x Long Holding Time: 22 Min #SOL #Solana #crypto #MaxQuantAI

Day 2 of MaxQuant AI Trading just began Entry price: 180.93 20x buy Exit price:181.84 holding time: 10 minutes Another day another win. Here we Go. #SOL #Trading #MaxQuantAI #crypto #CryptoMarket

MaxQuant AI | First Day Trading Summary 💰Profit: $48.86 💼Capital: $1,000 ⏱Duration: 24 hours 📈ROI: 4.89% Fully automated, signal-driven performance. Stay Tune for Day 2. #MaxQuantAI #AITrading #CryptoAI #BTC #Bitcoin #Trading