[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-1550":3,"related-1550":72},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":8,"published_at":11,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":23,"tags":25,"primary_topic":31,"secondary_topics":37,"classification":38,"search_phrases":66,"slug":69,"view_count":70,"doi":8,"paper":8,"created_at":71},1550,"中国农业大学王耀君团队\"神农育种 Studio\"9 月 4 日正式发布：1 套小麦育种方案效率提升 20 倍","https:\u002F\u002Fwww.beijing.gov.cn\u002Fywdt\u002Fgzdt\u002F202608\u002Ft20260831_4844377.html","9 月 4 日，\"神农育种 Studio\"正式发布，把原本分散的研究环节自动串联起来。育种人员用自然语言提出目标，Studio 先识别任务所需的数据、模型和软件，再按需调用相应工具，完成资料检索、亲本筛选、品系评价、多环境试验分析等任务。支撑平台运行的是团队多年积累的农业知识、生产数据和育种模型。神农大模型的数据底座已汇集超过 1000 万条农业知识图谱数据、超过 5000 万条农业生产数据。过去设计一套小麦育种方案需投入 10 人月，借助 Studio 可降至 0.5 人月，效率提升约 20 倍。",null,"田野里，无人车沿玉米行前行，机器一边“看”，算法一边“算”；鸡舍里，种鸡的产蛋、体重等数据不断汇入系统，基因芯片与算法共同寻找更有潜力的个体；实验室里，育种人员用自然语言提出目标，人工智能开始检索资料、调用模型、生成方案。\n\n这些正在北京发生的新场景，勾勒出同一个方向：育种正在从主要依靠经验，走向数据与经验结合、人脑与机器协同。\n\n9月4日至5日，第三十三届中国北京种业大会将在北京举行。大会以“AI赋能智慧育种·数智引领种业未来”为主题，首次以智慧育种为主线。人工智能怎样进入田间、鸡舍和实验室？它究竟改变了育种的哪一步？答案藏在正在采集的数据、实际的选种过程和不断落地的技术中。\n\n**机器“看”玉米**\n\n在北京通州区于家务乡的通州国际种业科技园区基地，20亩试验田里的约7万株玉米，正在接受一场“全周期体检”。环绕式无人车拍下植株的颜色、纹理和株型，激光雷达勾勒三维结构，多光谱和热成像设备捕捉长势与温度变化。一台不足半米宽的机器人则钻入冠层之下，从另一个视角补上从上方看不到的信息，近距离记录茎秆粗细、植株疏密和吐丝进度。\n\n从6叶期到9叶期，再到大喇叭口期和吐丝期，设备目前已完成4个阶段的数据采集，后续还将跟踪至成熟期。一次作业由两人操作，每小时可完成约4000株的数据采集。1000株以内的小批量数据，4至6小时可以形成结果，全区域数据可在24小时内完成解析和质控。在海南崖州南繁基地，同类装备还与无人机协同作业，在240亩试验田完成了超过36万株玉米的全生育期数据采集。\n\n交到育种人员手中的，不只是田间照片，还有性状数据表、三维点云模型、品系对比图表和单株表型档案。阶段性数据已能初步区分不同品系的株高增长速度、叶夹角和冠层透光性差异。团队随机抽取100株玉米，以人工测量的株高作对照，机器测量结果的一致性超过95%，误差低于5厘米。\n\n“过去依靠人工作业和经验判断，现在通过机器标准化采集、算法自动化处理，获取的指标更多、维度更多，精度也更高。”北京市农林科学院信息技术研究中心研究员郭新宇介绍，表型是基因型与环境互作的结果，只有把表型数据与基因组数据结合起来，才能更好地认识不同基因型在真实田间究竟会有怎样的表现。\n\n表型大数据工厂所做的，就是把作物真实的“长相”和“长势”变成AI可以学习的语言，帮助育种人员判断哪些材料在耐密、抗倒伏和高产方面更具潜力。\n\n**算法“评”种鸡**\n\n一只尚未产蛋的雏鸡，能不能提前判断它将来产多少蛋、蛋重多大、蛋壳结不结实？在北京市华都峪口禽业有限责任公司，一管血可以帮助育种人员更早读出候选种鸡的遗传潜力。\n\n研发人员闫奕源把选种比作选学生：传统方法要等母鸡产蛋，取得产蛋数、蛋重、蛋品质等“考试成绩”后才能判断。有了基因数据和预测模型，则可以在候选种鸡尚无自身生产记录时，提前评价其遗传潜能。\n\n雏鸡出生后采集血液、提取DNA，再用蛋鸡育种芯片检测基因变异位点。亲本和群体的历史数据中，既有基因型，也有产蛋数、蛋重、体重、蛋壳颜色和强度等表型数据。模型从两类数据的对应关系中学习，再对新一批候选种鸡逐只预测，形成涵盖不同性状的“成绩单”和综合排名。\n\n在这套体系中，芯片负责检测基因变异，算法模型则接过大规模比较、遗传潜能预测和排序等工作，帮助育种人员更早看出哪些个体更有育种潜力。“选得准，才能更快实现育种目标。”闫奕源说。\n\n目前，“京粉6号”已在百万只规模养殖条件下稳定实现“700天产500枚蛋”，相关成果获得2025年度国家科学技术进步奖二等奖。接续这一成熟品种，本届大会将发布“京红2号”“京粉8号”“沃德178”“京翠绿壳”等自主研发蛋（肉）鸡新品种，呈现“京字号”种鸡持续迭代的新成果。\n\n**平台“串”流程**\n\n如果说田间表型工厂解决的是“数据从哪里来”，种鸡精准育种展示的是“算法怎样辅助选择”，那么即将在本届大会发布的“神农育种Studio”，要解决的是分散的数据、模型和软件怎样被育种人员更方便地用起来。\n\n过去，育种人员常常需要在数据库、表格和多种软件间反复切换，有些环节还要请信息技术人员编写程序。现在，育种人员用自然语言提出目标，Studio先识别任务所需的数据、模型和软件，再按需调用相应工具，完成资料检索、亲本筛选、品系评价、多环境试验分析等任务。各环节自动衔接，前一步的结果成为下一步的输入，最后汇成附有资料来源的报告。\n\n支撑平台运行的，是团队多年积累的农业知识、生产数据和育种模型。神农大模型的数据底座已汇集超过1000万条农业知识图谱数据、超过5000万条农业生产数据。在北京市怀柔区，团队借助平台从数百个生菜品种中海选出“雁栖1号”“雁栖2号”。中国农业大学副教授王耀君介绍，过去一个蔬菜新品种往往要五六年才能培育出来，借助海量数据分析，可将时间缩短至两三年。\n\n9月4日，“神农育种Studio”将正式发布。Studio把原本分散的研究环节自动串联起来，效率变化随之显现。以设计一套小麦育种方案为例，王耀君介绍，过去需要投入10人月——即1个人工作10个月，或10个人工作1个月；在育种人员把目标和逻辑梳理清楚后，借助Studio可降至0.5人月。资料检索、数据处理、模型计算和初步方案生成等原本分散的工作，可由平台连续完成。\n\n从机器进田采数据，到算法辅助选材料，再到大模型串起研究环节，3种场景勾勒出智能育种的一条数据链路，也对算力、数据、装备和公共服务提出新要求。\n\n北京市农业农村局有关负责人介绍，北京将依托农业中关村核心区，打造北京农业中关村智能育种创新基地，计划建设智能育种共性技术创新中心和智能育种共性装备创新中心，提升育种专用算力、数据治理、装备研发中试、基因检测、标准规范和成果转化等共性条件；面向作物、畜禽、水产、微生物打造四类智能育种测试平台与产业应用场景。目前，作物逆境表型鉴定平台、作物智能分子育种创新平台、畜禽智能育种技术共性平台等重大平台正在建设中。\n\n育种仍是一场以年为单位的长跑，人工智能不能让种子跳过生长季，却能接过大量重复、耗时的前期工作。它可以把田间观察变成更完整的数据，把海量候选材料排出次序，把分散的研究环节自动串联起来，让每一轮选择少花一些时间、少走一些弯路。当机器真正读懂田间，数据、算法与育种经验形成合力，一粒好种子走向生产的脚步，便能迈得更快、更稳。","北京市人民政府","2026-09-04T00:00:00Z","报道",10,true,92,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":13,"relevant":21,"comment":22},28,24,18,12,1,"北京种业大会发布神农育种Studio，展示AI赋能育种全链条，效率提升20倍，具有产业级影响。",[24],{"name":10,"url":6},[26,27,28,29,30],"农业人工智能","北京","种业振兴","大模型","智慧育种",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},"A2","wheat","小麦","\u002Ftopic\u002Ffield-crops\u002Fwheat","A",[],{"status":39,"locked":40,"version":41,"method":42,"evidence":43},"classified",false,"2026-10-22.2","rules",{"candidates":44,"agricultural_evidence":14},[45,50,56,61],{"code":32,"score":46,"supported":14,"title":47,"abstract":48,"tags":49},11,[34],[34],[],{"code":51,"score":52,"supported":14,"title":53,"abstract":54,"tags":55},"I2",5,[],[29],[26,29],{"code":57,"score":21,"supported":40,"title":58,"abstract":59,"tags":60},"F1",[],[],[28],{"code":62,"score":21,"supported":40,"title":63,"abstract":64,"tags":65},"F2",[],[],[30],[67,68],"农业人工智能 智慧育种 种业振兴 大模型","农业人工智能 智慧育种","农业人工智能智慧育种种业振兴大模型-1550",3,"2026-09-04T00:05:30.233316Z",{"total":73,"page":21,"page_size":73,"items":74},6,[75,123,166,206,293,344],{"id":76,"title":77,"url":78,"summary":79,"summary_zh":8,"content":8,"source_name":80,"source_url":8,"published_at":81,"category":12,"cover_url":8,"hotness":13,"is_selected":40,"score":82,"score_detail":83,"sources":89,"tags":91,"primary_topic":95,"secondary_topics":96,"classification":97,"search_phrases":117,"slug":120,"view_count":121,"doi":8,"paper":8,"created_at":122},4859,"机器人变身小麦育种专家——山东农大小麦育种具身机器人原型机公开亮相","https:\u002F\u002Fnews.sciencenet.cn\u002Fsbhtmlnews\u002F2026\u002F10\u002F391582.shtm","近日在山东泰安举行的小麦智能育种研讨会上,小麦育种具身机器人原型机首次公开亮相。该机器人由山东农业大学教授王宏伟、张亮团队联合研发,采用实时动态差分定位与视觉融合导航技术,可在试验田小区间自主移动作业;搭载滑台高清相机,可对田间小麦进行连续影像采集,并利用视觉算法自动量化株高、穗部性状等表型参数。系统结合智能决策模型对田间单株进行高通量初筛,自主识别具有进一步考察价值的候选植株。研究团队与山东极智生物科技有限公司达成合作意向。","中国科学报 科学网 2026 年 10 月 13 日 第 1 版","2026-10-13T00:00:00Z",81,{"impact":84,"substance":85,"depth":86,"authority":87,"freshness":13,"relevant":21,"comment":88},22,20,16,13,"国内小麦育种具身机器人原型机首次公开亮相，技术路线与产学研合作信息具体，具备行业示范价值，值得进入每日精选。",[90],{"name":80,"url":78},[92,26,28,93,94],"智慧农业","小麦育种","具身机器人",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[],{"status":39,"locked":40,"version":41,"method":42,"evidence":98},{"candidates":99,"agricultural_evidence":14},[100,104,108,113],{"code":32,"score":20,"supported":14,"title":101,"abstract":102,"tags":103},[34],[34],[93],{"code":57,"score":21,"supported":40,"title":105,"abstract":106,"tags":107},[],[],[28],{"code":109,"score":21,"supported":40,"title":110,"abstract":111,"tags":112},"I1",[],[],[92],{"code":51,"score":21,"supported":40,"title":114,"abstract":115,"tags":116},[],[],[26],[118,119],"山东农业大学 小麦育种 具身机器人","王宏伟 小麦 表型采集","山东农业大学小麦育种具身机器人-4859",0,"2026-10-14T00:04:08.166418Z",{"id":124,"title":125,"url":126,"summary":127,"summary_zh":8,"content":128,"source_name":129,"source_url":8,"published_at":130,"category":12,"cover_url":8,"hotness":13,"is_selected":40,"score":131,"score_detail":132,"sources":136,"tags":138,"primary_topic":139,"secondary_topics":140,"classification":141,"search_phrases":161,"slug":164,"view_count":73,"doi":8,"paper":8,"created_at":165},4382,"山东农业大学研发小麦育种具身机器人 WheatX 首次亮相","http:\u002F\u002Fwww.shandong.gov.cn\u002Fart\u002F2026\u002F10\u002F6\u002Fart_97564_765520.html","山东农业大学张亮团队联合山东省农业科学院、中国农业大学等单位研发小麦育种具身机器人 WheatX,集感知、决策与选育功能于一体,可完成大田麦穗快速筛选、对量化达标麦穗喷涂颜料原位标记,实现「感知—决策—执行」全流程闭环;团队已与极智生物达成合作,将首款小麦育种大模型 iwheat 嵌入机器人,推动从原型机向产品级升级。","小麦试验田里，一台长宽高均超过两米的“大家伙”在麦株前稳稳停住。机械臂移动滑台探出，前端圆环装置精准套住一株麦穗缓缓上滑，环内八枚微型摄像头从多个角度对其精细成像。随后，穗粒数、穗长等性状的量化数据全部生成。\n\n这是由山东农业大学、小麦育种全国重点实验室王宏伟、张亮教授团队联合研发的小麦育种具身机器人原型机，近日在该校举办的小麦智能育种研讨会上首次公开亮相。\n\n王宏伟介绍，传统小麦育种中，育种家经年积累了大量宝贵育种经验，能从成千上万的杂交后代里选育出性状优良的品系。但在大田环境下，人工筛选不仅劳动强度大，筛选结果还容易受主观经验差异影响。同时，现有田间装备多以被动扫描和影像采集为主，能够完成“看”和“测”，但在自主判断、主动干预和选育执行等方面仍存在不足。该研究主要聚焦传统小麦育种主要依赖人工、选育效率低等行业共性问题，将为大田育种智能化提供新的装备与技术路径。\n\n“在模型训练时加入大量育种家先验经验是该款机器人特色之一。”王宏伟说，团队的目标，是让机器人不仅能够“看得见”，还能够“作判断、去操作”，在田间完成类似育种家的选育动作。\n\n为让机器人“学会”育种，团队种植了2000份不同的小麦种质资源，逐一拍摄影像后，请育种家按标准打分标注，再将标注结果输入系统，育种家凭多年经验作出的判断被转化为机器可学习的数据。机器人结合育种家经验构建智能决策模型，具备了初步的自主选育判断能力。\n\n同时，该机器人采用RTK与视觉融合导航技术，可在试验田小区间自主移动作业；搭载滑台高清相机，对田间小麦进行连续影像采集，并利用视觉算法自动量化株高、穗部性状等表型参数。在此基础上，系统结合智能决策模型，可对田间单株进行高通量初筛，自主识别具有进一步考察价值的候选植株。\n\n“对于初筛获得的目标植株，机器人还可通过自主研发的多功能机械臂开展进一步精细作业，包括单株抓取、环形多角度精细成像等操作。”张亮解释，对于量化后确认符合标准的麦穗，机器人会像育种家在田间“绑绳子做标记”一样，喷涂颜料原位标记，完成从大范围快速筛选到目标植株精细评价和标记的连续作业，实现“感知—决策—执行”的全流程闭环。\n\n该研究获山东省重点研发计划项目支持，并联合山东省农业科学院、中国农业大学等单位协同攻关。国家小麦产业技术体系副首席曹新有表示，该机器人集感知、决策与选育功能于一体，为智能育种由算法模型走向田间自主作业提供了关键技术支撑。\n\n“用机器人辅助人工选择，这是很难但很有意义的方向。”张亮坦言，大田场景麦穗密度高、类型多，场景极其复杂，加之天气影响，对机器人精准作业都是极大考验，系统在抓取精准度、学习经验深度等指标上还需进一步升级改良。目前，团队已与极智生物科技有限公司达成合作意向，将我国首款小麦育种大模型iwheat嵌入WheatX育种机器人，联合研发小麦智能育种系统，推动原型机迭代升级向产品级迈进。（记者 刘姝彤 通讯员 屈越）","山东省人民政府网 2026-10-06","2026-10-06T00:00:00Z",82,{"impact":84,"substance":84,"depth":133,"authority":87,"freshness":134,"relevant":21,"comment":135},17,8,"国内首款小麦育种具身机器人原型机亮相，实现感知—决策—执行闭环，属智慧育种领域实质性技术突破，值得进入每日精选。",[137],{"name":129,"url":126},[92,26,28,93,94],{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[],{"status":39,"locked":40,"version":41,"method":42,"evidence":142},{"candidates":143,"agricultural_evidence":14},[144,148,153,157],{"code":32,"score":20,"supported":14,"title":145,"abstract":146,"tags":147},[34],[34],[93],{"code":51,"score":149,"supported":14,"title":150,"abstract":151,"tags":152},4,[],[29],[26],{"code":57,"score":21,"supported":40,"title":154,"abstract":155,"tags":156},[],[],[28],{"code":109,"score":21,"supported":40,"title":158,"abstract":159,"tags":160},[],[],[92],[162,163],"山东农业大学 小麦育种机器人 WheatX","小麦智能育种 iwheat 大模型","山东农业大学小麦育种机器人WheatX-4382","2026-10-08T00:06:07.051280Z",{"id":167,"title":168,"url":169,"summary":170,"summary_zh":8,"content":171,"source_name":172,"source_url":8,"published_at":173,"category":12,"cover_url":8,"hotness":13,"is_selected":40,"score":174,"score_detail":175,"sources":177,"tags":179,"primary_topic":180,"secondary_topics":181,"classification":182,"search_phrases":201,"slug":203,"view_count":204,"doi":8,"paper":8,"created_at":205},4262,"山东农业大学发布小麦育种具身机器人原型机","https:\u002F\u002Fwww.toutiao.com\u002Farticle\u002F7692412028073361958","山东农业大学小麦育种全国重点实验室王宏伟、张亮教授团队联合山东农科院、中国农业大学、北京大学现代农业研究院研发的小麦育种具身机器人原型机公开亮相。机器人采用RTK与视觉融合导航，可自主移动作业，搭载滑台高清相机连续影像采集，结合育种家先验经验构建智能决策模型，对田间单株高通量初筛与精细作业。","## 能判断，懂操作！小麦育种具身机器人原型机公开亮相\n\n近日，在山东农业大学举办的小麦智能育种研讨会上，由山东农业大学小麦育种全国重点实验室王宏伟、张亮教授团队联合研发的小麦育种具身机器人原型机首次公开亮相。该研究获山东省重点研发计划项目（2026LZGC001）支持，并联合山东农业科学院、中国农业大学、北京大学现代农业研究院等单位协同攻关，聚焦传统小麦育种中田间表型鉴定效率低、单株筛选依赖人工等行业共性问题，为田间育种智能化提供新的装备与技术路径。\n\n视频加载中...\n\n长期以来，小麦田间表型鉴定和单株筛选高度依赖人工，不仅作业通量有限、劳动强度较大，筛选结果还容易受育种人员经验差异的影响。与此同时，现有田间表型采集装备多以被动扫描和影像采集为主，能够完成“看”和“测”，但在自主判断、主动干预和选育执行等方面仍存在不足。\n\n针对这些问题，研发团队研制了面向大田小麦育种的高通量主动式机器人系统，并以“主动感知—自主决策—联动执行”为核心设计理念，推动田间表型装备由单纯数据采集向自主选育作业拓展。该机器人采用RTK与视觉融合导航技术，可在试验田小区间自主移动作业；搭载滑台高清相机，对田间小麦进行连续影像采集，并利用视觉算法自动量化株高、穗部性状等表型参数。在此基础上，系统结合育种家先验经验构建智能决策模型，可对田间单株进行高通量初筛，自主识别具有进一步考察价值的候选植株。对于初筛获得的目标植株，机器人还可通过自主研发的机械臂开展进一步精细作业，包括单株抓取、环形多角度精细成像以及原位喷涂标记等操作，实现从大范围快速筛选到目标植株精细评价和标记的连续作业，使机器人不仅能够“看得见”，还能够“做判断、去操作”。\n\n研发团队成员、山东农业大学教授王宏伟介绍：“整套系统突破了现有表型装备以被动扫描为主、缺乏主动干预能力的技术局限，打通了小麦育种机器‘感知—决策—执行’的全流程闭环。”国家小麦产业技术体系副首席曹新有表示，此次发布的机器人集感知、决策与选育功能于一体，为智能育种由算法模型走向田间自主作业提供了关键技术支撑。\n\n此次小麦智能育种研讨会上，来自中国农业科学院、中国农业大学、北京大学现代农业研究院等单位的专家还围绕种质资源数据整合、人工智能算法共享和育种智能装备研发等开展深入交流，并达成进一步合作意向。各方将围绕小麦育种关键环节持续推进数据、算法与智能装备协同创新，探索构建覆盖“感知—分析—决策—执行”的智能育种技术体系，推动小麦育种向数字化、智能化和自主化方向发展。\n\n（大众新闻记者 刘姝彤 通讯员 王宏伟）","今日头条 大众日报 2026-10-03","2026-10-03T12:07:00Z",74,{"impact":84,"substance":85,"depth":86,"authority":13,"freshness":73,"relevant":21,"comment":176},"国内首台小麦育种具身机器人原型机亮相，打通感知—决策—执行闭环，属智能育种装备领域实质性突破，值得进入每日精选。",[178],{"name":172,"url":169},[92,26,28,93,94],{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[],{"status":39,"locked":40,"version":41,"method":42,"evidence":183},{"candidates":184,"agricultural_evidence":14},[185,189,193,197],{"code":32,"score":20,"supported":14,"title":186,"abstract":187,"tags":188},[34],[34],[93],{"code":57,"score":21,"supported":40,"title":190,"abstract":191,"tags":192},[],[],[28],{"code":109,"score":21,"supported":40,"title":194,"abstract":195,"tags":196},[],[],[92],{"code":51,"score":21,"supported":40,"title":198,"abstract":199,"tags":200},[],[],[26],[118,202],"小麦育种全国重点实验室 智能育种","山东农业大学小麦育种具身机器人-4262",2,"2026-10-06T00:05:02.013148Z",{"id":207,"title":208,"url":209,"summary":210,"summary_zh":211,"content":8,"source_name":212,"source_url":209,"published_at":213,"category":214,"cover_url":8,"hotness":13,"is_selected":40,"score":82,"score_detail":215,"sources":217,"tags":219,"primary_topic":223,"secondary_topics":224,"classification":236,"search_phrases":265,"slug":268,"view_count":149,"doi":269,"paper":270,"created_at":292},3263,"Identifying drought-responsive candidate genes in root and shoot tissues of wheat (Triticum aestivum L.) through RNA sequencing, cluster analysis, weighted gene co-expression network analysis, and machine learning models","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffsufs.2026.1952456","Drought stress severely affects the growth, development, and yield formation of wheat. However, the differentiated regulatory mechanisms of wheat roots and shoots during drought adaptation remain to be further analyzed. Therefore, this study used RNA-seq datasets of roots and shoots from multiple wheat varieties, combined with differential expression analysis, Gene Ontology (GO) enrichment analysis, clustering analysis, weighted gene co-expression network analysis (WGCNA), and machine learning were used to characterize drought-responsive transcriptional changes. Differential expression analysis identified 90 and 3,991 differentially expressed genes (DEGs) in roots under DS1 and DS2 respectively, compared with 34 and 1,940 DEGs in shoots. Functional enrichment analysis indicated that root DEGs were mainly involved in root growth, cell wall remodeling, water transport, and osmotic adjustment processes, whereas shoot DEGs are mainly enriched in processes related to stomatal regulation, photosynthesis, lipid metabolism, and oxidative stress responses. WGCNA identified 18 and 19 co-expression modules in roots and shoots, respectively, among which the yellow module was significantly correlated with the response to drought stress in root and shoot tissues. To further screen drought-responsive candidate genes from these modules, we compared multiple machine learning models. The XGBoost model achieved the highest and AUC in both roots and shoots. Combined with SHAP feature contribution analysis, 30 drought-responsive candidate genes with high discriminatory power were identified separately in roots and shoots. Analysis of cis -acting elements in candidate gene promoters revealed abundant drought- and stress-responsive elements, including MYB, G-box, ABRE, MYC, MBS, and LTR. MYB (183 in roots and 164 in shoots), G-box (168 and 98), and ABRE (148 and 86) were particularly abundant, and several candidate genes contained more than five ABRE copies in their promoters, suggesting potential involvement in ABA-mediated drought responses. This study reveals the different transcriptional response processes of wheat roots and shoots under drought stress, and identifies key co-expression modules and core candidate genes through WGCNA combined with XGBoost and SHAP analysis. These findings provide a basis for understanding organ-specific molecular responses to drought stress and offer potential targets for improving drought tolerance in wheat.","干旱胁迫严重影响小麦的生长发育和产量形成。然而，小麦根系和地上部在干旱适应过程中的差异化调控机制仍有待深入解析。因此，本研究利用多个小麦品种根系和地上部的RNA-seq数据集，结合差异表达分析、基因本体（GO）富集分析、聚类分析、加权基因共表达网络分析（WGCNA）和机器学习方法，表征干旱响应转录变化。差异表达分析在DS1和DS2条件下分别鉴定出根系中90个和3,991个差异表达基因（DEGs），而地上部中分别为34个和1,940个DEGs。功能富集分析表明，根系DEGs主要参与根系生长、细胞壁重塑、水分运输和渗透调节过程，而地上部DEGs主要富集于气孔调控、光合作用、脂质代谢和氧化胁迫响应相关过程。WGCNA在根系和地上部中分别鉴定出18个和19个共表达模块，其中黄色模块与根系和地上部组织对干旱胁迫的响应显著相关。为进一步从这些模块中筛选干旱响应候选基因，我们比较了多种机器学习模型。XGBoost模型在根系和地上部中均取得了最高的AUC。结合SHAP特征贡献分析，分别在根系和地上部中鉴定出30个具有高判别力的干旱响应候选基因。候选基因启动子中的顺式作用元件分析揭示了丰富的干旱和胁迫响应元件，包括MYB、G-box、ABRE、MYC、MBS和LTR。其中MYB（根系183个，地上部164个）、G-box（168个和98个）和ABRE（148个和86个）尤为丰富，且多个候选基因的启动子中含有超过5个ABRE拷贝，表明其可能参与ABA介导的干旱响应。本研究揭示了小麦根系和地上部在干旱胁迫下不同的转录响应过程，并通过WGCNA结合XGBoost和SHAP分析鉴定了关键共表达模块和核心候选基因。这些发现为理解干旱胁迫的器官特异性分子响应提供了基础，并为提高小麦耐旱性提供了潜在靶点。","Frontiers in Sustainable Food Systems","2026-09-23T00:00:00Z","论文",{"impact":19,"substance":84,"depth":19,"authority":87,"freshness":13,"relevant":21,"comment":216},"该研究结合多组学与机器学习挖掘小麦抗旱候选基因，方法新颖、数据扎实，对分子育种有参考价值，值得进入每日精选。",[218],{"name":212,"url":209},[26,28,220,221,222],"基因挖掘","作物育种","小麦抗旱",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[225,230],{"code":51,"slug":226,"name":227,"path":228,"parent_code":229},"agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",{"code":231,"slug":232,"name":233,"path":234,"parent_code":235},"H7","climate","气候适应与防灾减灾","\u002Ftopic\u002Fenvironment\u002Fclimate","H",{"status":39,"locked":40,"version":41,"method":42,"evidence":237},{"candidates":238,"agricultural_evidence":14},[239,246,251,257,261],{"code":51,"score":240,"supported":14,"title":241,"abstract":243,"tags":245},15,[242],"machine learning",[244,242],"机器学习",[26],{"code":32,"score":247,"supported":14,"title":248,"abstract":249,"tags":250},14,[33],[34,33],[],{"code":231,"score":247,"supported":14,"title":252,"abstract":254,"tags":256},[253],"drought",[255,253],"干旱",[],{"code":57,"score":21,"supported":40,"title":258,"abstract":259,"tags":260},[],[],[28],{"code":62,"score":21,"supported":40,"title":262,"abstract":263,"tags":264},[],[],[220],[266,267],"小麦 抗旱基因 RNA测序","小麦 根系 地上部 干旱响应","小麦抗旱基因RNA测序-3263","10.3389\u002Ffsufs.2026.1952456",{"doi":269,"openalex_id":271,"authors":272,"venue":212,"cited_by_count":121,"oa_url":209,"card":285,"direction":289,"ingested_from":291},"W7214062209",[273,275,278,280,282],{"name":274,"orcid":8},"Tianle Ji",{"name":276,"orcid":277},"Baoyue Cui","https:\u002F\u002Forcid.org\u002F0009-0000-5881-9682",{"name":279,"orcid":8},"Jinjie Yin",{"name":281,"orcid":8},"Xianyong Meng",{"name":283,"orcid":284},"Jun Yan","https:\u002F\u002Forcid.org\u002F0000-0001-6883-8529",{"tldr":286,"method":287,"finding":288,"direction":289,"opportunity":290},"通过RNA-seq、WGCNA和XGBoost-SHAP分析，鉴定小麦根和茎中响应干旱的候选基因。","RNA-seq、差异表达、WGCNA、XGBoost与SHAP特征分析。","根和茎干旱响应机制不同，各鉴定出30个候选基因，启动子富含MYB、G-box和ABRE元件。","农业人工智能与决策模型","可结合多组学与深度学习构建器官特异性干旱调控网络，并开展候选基因功能验证。","openalex","2026-09-23T23:30:07.062126Z",{"id":294,"title":295,"url":296,"summary":297,"summary_zh":8,"content":298,"source_name":299,"source_url":8,"published_at":300,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":301,"score_detail":302,"sources":304,"tags":306,"primary_topic":308,"secondary_topics":309,"classification":310,"search_phrases":339,"slug":342,"view_count":149,"doi":8,"paper":8,"created_at":343},1465,"中国农业大学王耀君团队\"神农育种 Studio\"将于 9 月 4 日发布：1 套小麦育种方案效率提升 20 倍","https:\u002F\u002Fnews.cau.edu.cn\u002Fmtndnew\u002F021c96341130432394f118cd42f2f462.htm","中国农业大学副教授王耀君介绍：神农大模型数据底座已汇集超过 1000 万条农业知识图谱数据、超过 5000 万条农业生产数据；团队借助平台从数百个生菜品种中海选出\"雁栖 1 号\"\"雁栖 2 号\"。9 月 4 日\"神农育种 Studio\"将正式发布——以设计一套小麦育种方案为例，过去需要投入 10 人月，借助 Studio 可降至 0.5 人月，资料检索、数据处理、模型计算和初步方案生成等原分散工作可由平台连续完成。北京将依托农业中关村核心区打造北京农业中关村智能育种创新基地。","![Image 1: 微信截图_20260902172646.png](https:\u002F\u002Fnews.cau.edu.cn\u002Fimages\u002F2026-09\u002Fe4600f17fe5c4b24885135b75827105b.png)\n\n**农民日报全媒体记者 芦晓春 赵天浩**\n\n田野里，无人车沿玉米行前行，机器一边“看”，算法一边“算”；鸡舍里，种鸡的产蛋、体重等数据不断汇入系统，基因芯片与算法共同寻找更有潜力的个体；实验室里，育种人员用自然语言提出目标，人工智能开始检索资料、调用模型、生成方案。\n\n这些正在北京发生的新场景，勾勒出同一个方向：育种正在从主要依靠经验，走向数据与经验结合、人脑与机器协同。\n\n9月4日至5日，第三十三届中国北京种业大会将在北京举行。大会以“AI赋能智慧育种·数智引领种业未来”为主题，首次以智慧育种为主线。人工智能怎样进入田间、鸡舍和实验室？它究竟改变了育种的哪一步？答案藏在正在采集的数据、实际的选种过程和不断落地的技术中。\n\n**机器“看”玉米**\n\n在北京通州区于家务乡的通州国际种业科技园区基地，20亩试验田里的约7万株玉米，正在接受一场“全周期体检”。环绕式无人车拍下植株的颜色、纹理和株型，激光雷达勾勒三维结构，多光谱和热成像设备捕捉长势与温度变化。一台不足半米宽的机器人则钻入冠层之下，从另一个视角补上从上方看不到的信息，近距离记录茎秆粗细、植株疏密和吐丝进度。\n\n从6叶期到9叶期，再到大喇叭口期和吐丝期，设备目前已完成4个阶段的数据采集，后续还将跟踪至成熟期。一次作业由两人操作，每小时可完成约4000株的数据采集。1000株以内的小批量数据，4至6小时可以形成结果，全区域数据可在24小时内完成解析和质控。在海南崖州南繁基地，同类装备还与无人机协同作业，在240亩试验田完成了超过36万株玉米的全生育期数据采集。\n\n交到育种人员手中的，不只是田间照片，还有性状数据表、三维点云模型、品系对比图表和单株表型档案。阶段性数据已能初步区分不同品系的株高增长速度、叶夹角和冠层透光性差异。团队随机抽取100株玉米，以人工测量的株高作对照，机器测量结果的一致性超过95%，误差低于5厘米。\n\n“过去依靠人工作业和经验判断，现在通过机器标准化采集、算法自动化处理，获取的指标更多、维度更多，精度也更高。”北京市农林科学院信息技术研究中心研究员郭新宇介绍，表型是基因型与环境互作的结果，只有把表型数据与基因组数据结合起来，才能更好地认识不同基因型在真实田间究竟会有怎样的表现。\n\n表型大数据工厂所做的，就是把作物真实的“长相”和“长势”变成AI可以学习的语言，帮助育种人员判断哪些材料在耐密、抗倒伏和高产方面更具潜力。\n\n算法“评”种鸡\n\n一只尚未产蛋的雏鸡，能不能提前判断它将来产多少蛋、蛋重多大、蛋壳结不结实？在北京市华都峪口禽业有限责任公司，一管血可以帮助育种人员更早读出候选种鸡的遗传潜力。\n\n研发人员闫奕源把选种比作选学生：传统方法要等母鸡产蛋，取得产蛋数、蛋重、蛋品质等“考试成绩”后才能判断。有了基因数据和预测模型，则可以在候选种鸡尚无自身生产记录时，提前评价其遗传潜能。\n\n雏鸡出生后采集血液、提取DNA，再用蛋鸡育种芯片检测基因变异位点。亲本和群体的历史数据中，既有基因型，也有产蛋数、蛋重、体重、蛋壳颜色和强度等表型数据。模型从两类数据的对应关系中学习，再对新一批候选种鸡逐只预测，形成涵盖不同性状的“成绩单”和综合排名。\n\n在这套体系中，芯片负责检测基因变异，算法模型则接过大规模比较、遗传潜能预测和排序等工作，帮助育种人员更早看出哪些个体更有育种潜力。“选得准，才能更快实现育种目标。”闫奕源说。\n\n目前，“京粉6号”已在百万只规模养殖条件下稳定实现“700天产500枚蛋”，相关成果获得2025年度国家科学技术进步奖二等奖。接续这一成熟品种，本届大会将发布“京红2号”“京粉8号”“沃德178”“京翠绿壳”等自主研发蛋（肉）鸡新品种，呈现“京字号”种鸡持续迭代的新成果。\n\n**平台“串”流程**\n\n**如果说田间表型工厂解决的是“数据从哪里来”，种鸡精准育种展示的是“算法怎样辅助选择”，那么即将在本届大会发布的“神农育种Studio”，要解决的是分散的数据、模型和软件怎样被育种人员更方便地用起来。**\n\n**过去，育种人员常常需要在数据库、表格和多种软件间反复切换，有些环节还要请信息技术人员编写程序。现在，育种人员用自然语言提出目标，Studio先识别任务所需的数据、模型和软件，再按需调用相应工具，完成资料检索、亲本筛选、品系评价、多环境试验分析等任务。各环节自动衔接，前一步的结果成为下一步的输入，最后汇成附有资料来源的报告。**\n\n**支撑平台运行的，是团队多年积累的农业知识、生产数据和育种模型。神农大模型的数据底座已汇集超过1000万条农业知识图谱数据、超过5000万条农业生产数据。在北京市怀柔区，团队借助平台从数百个生菜品种中海选出“雁栖1号”“雁栖2号”。中国农业大学副教授王耀君介绍，过去一个蔬菜新品种往往要五六年才能培育出来，借助海量数据分析，可将时间缩短至两三年。**\n\n**9月4日，“神农育种Studio”将正式发布。Studio把原本分散的研究环节自动串联起来，效率变化随之显现。以设计一套小麦育种方案为例，王耀君介绍，过去需要投入10人月——即1个人工作10个月，或10个人工作1个月；在育种人员把目标和逻辑梳理清楚后，借助Studio可降至0.5人月。资料检索、数据处理、模型计算和初步方案生成等原本分散的工作，可由平台连续完成。**\n\n**从机器进田采数据，到算法辅助选材料，再到大模型串起研究环节，3种场景勾勒出智能育种的一条数据链路，也对算力、数据、装备和公共服务提出新要求。**\n\n北京市农业农村局有关负责人介绍，北京将依托农业中关村核心区，打造北京农业中关村智能育种创新基地，计划建设智能育种共性技术创新中心和智能育种共性装备创新中心，提升育种专用算力、数据治理、装备研发中试、基因检测、标准规范和成果转化等共性条件；面向作物、畜禽、水产、微生物打造四类智能育种测试平台与产业应用场景。目前，作物逆境表型鉴定平台、作物智能分子育种创新平台、畜禽智能育种技术共性平台等重大平台正在建设中。\n\n育种仍是一场以年为单位的长跑，人工智能不能让种子跳过生长季，却能接过大量重复、耗时的前期工作。它可以把田间观察变成更完整的数据，把海量候选材料排出次序，把分散的研究环节自动串联起来，让每一轮选择少花一些时间、少走一些弯路。当机器真正读懂田间，数据、算法与育种经验形成合力，一粒好种子走向生产的脚步，便能迈得更快、更稳。\n\n[《农民日报》2026年8月28日 1版](https:\u002F\u002Fszb.farmer.com.cn\u002Fnmrb\u002Fhtml\u002F2026\u002F20260828\u002F20260828_1\u002Fnmrb_20260828_13390_1_2093084220427702292.html)","中国农业大学新闻网","2026-08-28T00:00:00Z",90,{"impact":17,"substance":18,"depth":19,"authority":247,"freshness":73,"relevant":21,"comment":303},"报道详实，展示AI在育种全链条的应用，发布平台效率提升显著，具产业级影响。",[305],{"name":299,"url":296},[92,26,28,307,93],"智能育种",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[],{"status":39,"locked":40,"version":41,"method":42,"evidence":311},{"candidates":312,"agricultural_evidence":14},[313,317,321,327,331,335],{"code":32,"score":20,"supported":14,"title":314,"abstract":315,"tags":316},[34],[34],[93],{"code":51,"score":149,"supported":14,"title":318,"abstract":319,"tags":320},[],[29],[26],{"code":322,"score":70,"supported":14,"title":323,"abstract":324,"tags":326},"B1",[],[325],"生菜",[],{"code":57,"score":21,"supported":40,"title":328,"abstract":329,"tags":330},[],[],[28],{"code":62,"score":21,"supported":40,"title":332,"abstract":333,"tags":334},[],[],[307],{"code":109,"score":21,"supported":40,"title":336,"abstract":337,"tags":338},[],[],[92],[340,341],"农业人工智能 小麦育种 智慧农业 智能育种","农业人工智能 小麦育种","农业人工智能小麦育种智慧农业智能育种-1465","2026-09-03T00:06:40.464354Z",{"id":345,"title":346,"url":347,"summary":348,"summary_zh":8,"content":8,"source_name":349,"source_url":8,"published_at":350,"category":12,"cover_url":8,"hotness":13,"is_selected":40,"score":351,"score_detail":352,"sources":354,"tags":356,"primary_topic":360,"secondary_topics":361,"classification":362,"search_phrases":379,"slug":382,"view_count":121,"doi":8,"paper":8,"created_at":383},5443,"四川省农科院作物所优质弱筋小麦新品种「川麦632」品种权转让公示","https:\u002F\u002Fwww.chinawestagr.com\u002Fhomepage\u002FshowAffiche.asp?id=3026","四川省农业科学院作物研究所（四川省种质资源中心）公示小麦品种「川麦632」品种权转让结果，受让方为四川均隆生物科技有限公司。「川麦632」是该所选育的优质弱筋小麦新品种，已于 2026 年 10 月通过品种报审，由蒲宗君等 9 位育种人完成。转让费采用入门费加提成费方式，入门费 30 万元，签订合同后支付 5 万元、审定公告出号后支付尾款 25 万元；提成费自 2029 年起按每销售一公斤种子提成 0.1 元，公示期至 2026 年 11 月 2 日。","四川省农业科学院 2026-10-19","2026-10-19T00:00:00Z",67,{"impact":86,"substance":85,"depth":20,"authority":87,"freshness":73,"relevant":21,"comment":353},"省级科研机构小麦新品种品种权转让公示，含转让费与提成条款等实质信息，对种业成果转化有参考价值，但属细分行业进展。",[355],{"name":349,"url":347},[28,357,34,358,359],"成果转化","品种权转让","弱筋小麦",{"code":32,"slug":33,"name":34,"path":35,"parent_code":36},[],{"status":39,"locked":40,"version":41,"method":42,"evidence":363},{"candidates":364,"agricultural_evidence":14},[365,369,374],{"code":32,"score":20,"supported":14,"title":366,"abstract":367,"tags":368},[34],[34],[34],{"code":57,"score":149,"supported":14,"title":370,"abstract":371,"tags":373},[],[372],"种质资源",[28],{"code":375,"score":21,"supported":40,"title":376,"abstract":377,"tags":378},"G1",[],[],[357],[380,381],"川麦632 品种权转让","四川省农科院 弱筋小麦","川麦632品种权转让-5443","2026-10-21T00:10:01.857857Z"]