Analyzing trait-climate relationships within and among taxa using machine learning and herbarium specimens

Premise

Continental-scale leaf trait studies can help explain how plants survive in different environments, but large data sets are costly to assemble at this scale. Automating the measurement of digitized herbarium collections could rapidly expand the data available to such studies. We used machine learning to identify and measure leaves from existing, digitized herbarium specimens. The process was developed, validated, and applied to analyses of relationships between leaf size and climate within and among species for two genera: Syzygium (Myrtaceae) and Ficus (Moraceae).

Methods

Convolutional neural network (CNN) models were used to detect and measure complete leaves in images. Predictions of a model trained with a set of 35 randomly selected images and a second model trained with 35 user-selected images were compared using a set of 50 labeled validation images. The validated models were then applied to 1227 Syzygium and 2595 Ficus specimens digitized by the National Herbarium of New South Wales, Australia. Leaf area measurements were made for each genus and used to examine links between leaf size and climate.

Results

The user-selected training method for Syzygium found more leaves (9347 vs. 8423) using fewer training masks (218 vs. 225), and found leaves with a greater range of sizes than the random image training method. Within each genus, leaf size was positively associated with temperature and rainfall, consistent with previous observations. However, within species, the associations between leaf size and environmental variables were weaker.

Conclusions

CNNs detected and measured leaves with levels of accuracy useful for trait extraction and analysis and illustrate the potential for machine learning of herbarium specimens to massively increase global leaf trait data sets. Within-species relationships were weak, suggesting that population history and gene flow have a strong effect at this level. Herbarium specimens and machine learning could expand sampling of trait data within many species, offering new insights into trait evolution.     

Changes in plant collection practices from the 16th to 21st centuries: implications for the use of herbarium specimens in global change research

Background and AimsHerbaria were recently advertised as reliable sources of information regarding historical changes in plant traits and biotic interactions. To justify the use of herbaria in global change research, we asked whether the characteristics of herbarium specimens have changed during the past centuries and whether these changes were due to shifts in plant collection practices.MethodsWe measured nine characteristics from 515 herbarium specimens of common European trees and large shrubs collected from 1558 to 2016. We asked botanists to rank these specimens by their scientific quality, and asked artists to rank these specimens by their beauty.Key ResultsEight of 11 assessed characteristics of herbarium specimens changed significantly during the study period. The average number of leaves in plant specimens increased 3-fold, whereas the quality of specimen preparation decreased. Leaf size negatively correlated with leaf number in specimens in both among-species and within-species analyses. The proportion of herbarium sheets containing plant reproductive structures peaked in the 1850s. The scientific value of herbarium specimens increased until the 1700s, but then did not change, whereas their aesthetic value showed no systematic trends.ConclusionsOur findings strongly support the hypothesis that many characteristics of herbarium specimens have changed systematically and substantially from the 16th to 21st centuries due to changes in plant collection and preservation practices. These changes may both create patterns which could be erroneously attributed to environmental changes and obscure historical trends in plant traits. The utmost care ought to be taken to guard against the possibility of misinterpretation of data obtained from herbarium specimens. We recommend that directional changes in characters of herbarium specimens which occurred during the past 150‒200 years, primarily in specimen size and in the presence of reproductive structures, are accounted for when searching for the effects of past environmental changes on plant traits.     

Amplification of DNA of Xanthomonas axonopodis pv. citri from historic citrus canker herbarium specimens

Herbaria are important resources for the study of the origins and dispersal of plant pathogens, particularly bacterial plant pathogens that incite local lesions in which large numbers of pathogen genomes are concentrated. Xanthomonas axonopodis pv. citri (Xac), the causal agent of citrus bacterial canker disease, is a notable example of such a pathogen. The appearance of novel strains of the pathogen in Florida and elsewhere make it increasingly important to understand the relationships among strains of this pathogen. USDA-ARS at Beltsville, Maryland maintains approximately 700 herbarium specimens with citrus canker disease lesions up to 90 years old, originally collected from all over the world, and so is an important resource for phytogeographic studies of this bacterium. Unfortunately, DNA in herbarium specimens is degraded and may contain high levels of inhibitors of PCR. In this study, we compared a total of 23 DNA isolation techniques in combination with 31 novel primer pairs in order to develop an efficient protocol for the analysis of Xac DNA in herbarium specimens. We identified the most reliable extraction method, identified in terms of successful amplification by our panel of 31 primer pairs. We also identified the most robust primer pairs, identified as successful in the largest number of extracts prepared by different methods. We amplified Xac genomic sequences up to 542 bp long from herbarium samples up to 89 years old. Primers varied in effectiveness, with some primer pairs amplifying Xac DNA from a 1/10,000 dilution of extract from a single lesion from a citrus canker herbarium specimen. Our methodology will be useful to identify pathogens and perform molecular analyses of bacterial and possibly fungal genomes from herbarium specimens.     

中国竹类植物馆藏标本现状与地理分布

馆藏标本是分类学研究的主要凭证, 对特定类群标本的采集信息进行细致整理和分析, 有助于理解该类群研究的历史、现状和不足。此外, 结合最新的系统学研究成果和相应的环境数据构建生态位模型, 可以加深人们对特定类群分类与分布状况的认识。被子植物的分类鉴定常基于繁殖性状进行, 然而, 竹类植物一般进行克隆繁殖, 只有在经过长期的营养生长之后, 才会进行有性生殖并同时死亡。因此, 国内的竹类标本大多仅记录了营养性状, 具有繁殖性状的标本数量稀少。由于这一特殊的生物学习性, 竹类植物是当今分类学研究中最为困难的类群之一。本研究基于全国竹类植物馆藏标本的采集数据, 分析了我国竹类标本的采集和保藏现状; 利用比值法和斜率法从采集地理偏差和采集类群偏差两方面评估了竹类植物标本的采集完整度; 结合气候数据, 利用模型模拟的方法分析了影响不同竹类分支分布的主要因素。采集信息分析结果表明, 国内标本馆对竹类标本的收集和保藏存在很大的不均衡性, 且对研究团队具有较高的依赖性; 其次竹类标本的采集量的变化较好地反映了国内植物分类学研究的历史; 而对类群和地理采集完整度的评估结果表明, 中国竹类标本的采集和整理工作仍任重道远。模型模拟结果表明, 温度限定了竹类植物两大分支各自的分布北界, 水分对温带木本竹类分支(temperate woody bamboos, TWB)的限制作用比旧世界热带木本竹类分支(paleotropical woody bamboos, PWB)强, 而温度对PWB的限制性更强。生态位模拟的结果进一步显示, 中国温带和旧世界热带木本竹类两大分支的适生区出现了明显的分化, 但在亚热带区域仍有部分重叠。植物标本记录了特定类群在时间和空间上的分布格局, 相关的信息一方面可以促进物种灭绝风险评估、可持续利用和综合保护, 另一方面也可助力大尺度生物多样性分布格局及全球变化对多样性的影响研究。     

A deep learning-based approach for detecting plant organs from digitized herbarium specimen images

Herbarium specimens are excellent sources of botanical information to facilitate understanding and monitoring the evolution of plants and their effects on global climate change. Globally, many herbaria have undertaken digitization projects of herbarium specimens to preserve them and make them accessible in online repositories to botanists and ecologists. Automated detection of plant organs such as plant leaves, buds, flowers, and fruits on the digitized herbarium specimen images provides valuable information in various scientific contexts. We developed a deep learning approach based on the refined YOLO-V3 approach to detect plant organs within the digitized herbarium specimen images effectively. The proposed approach combines ResNet and DenseNet architectures to improve feature extraction capabilities. Also, a new scale of feature map is added to the existing scales to address the problem of YOLO-V3's low performance in detecting small plant organs. The experimental results demonstrate that our proposed approach can detect organs of different sizes within different specimens, where the precision and recall reached 94.2% and 95.5%, respectively.     

植物标本标签设计的原则及R程序包herblabel

植物标本是分类学、生态学和分子生物学最重要的凭证之一。标本的采集和鉴定信息需清晰、准确、美观地展示和保存于标本标签中, 不能有歧义以及拼写错误。在标签的制作过程中, 数据输入的方式要简单、直接, 标签文件生成过程中最好能自动分析错误, 且在打印之前要便于修改和调整。本文探讨了打印植物标本标签的若干原则以及注意事项, 并介绍了用R语言编写的herblabel程序包生成植物标本标签以及鉴定标签的过程。herblabel程序包基于Darwin Core和CVH5.0数据交换标准, 可快速批量生成几种样式的RTF标签, 且标签简洁、美观, 易于编辑。herblabel程序包具有检查地点完整程度, 学名拼写和接受状态, 科、属在APG等新系统下的对应关系等功能, 可有效减少数据录入过程中产生的错误。此外, 本程序包在打印标签时使用的是基于Darwin Core标准保存的标本数据库, 不仅方便统计和管理, 也可以直接用于全球生物多样性信息网络(GBIF)数据共享或者数字植物标本馆的建设。该程序包可显著提高植物标本馆标本制作、管理和信息录入的工作效率, 减轻工作人员的负担, 并在植物生物多样性编目中发挥重要作用。     

DNA damage in plant herbarium tissue

Dried plant herbarium specimens are potentially a valuable source of DNA. Efforts to obtain genetic information from this source are often hindered by an inability to obtain amplifiable DNA as herbarium DNA is typically highly degraded. DNA post-mortem damage may not only reduce the number of amplifiable template molecules, but may also lead to the generation of erroneous sequence information. A qualitative and quantitative assessment of DNA post-mortem damage is essential to determine the accuracy of molecular data from herbarium specimens. In this study we present an assessment of DNA damage as miscoding lesions in herbarium specimens using 454-sequencing of amplicons derived from plastid, mitochondrial, and nuclear DNA. In addition, we assess DNA degradation as a result of strand breaks and other types of polymerase non-bypassable damage by quantitative real-time PCR. Comparing four pairs of fresh and herbarium specimens of the same individuals we quantitatively assess post-mortem DNA damage, directly after specimen preparation, as well as after long-term herbarium storage. After specimen preparation we estimate the proportion of gene copy numbers of plastid, mitochondrial, and nuclear DNA to be 2.4-3.8% of fresh control DNA and 1.0-1.3% after long-term herbarium storage, indicating that nearly all DNA damage occurs on specimen preparation. In addition, there is no evidence of preferential degradation of organelle versus nuclear genomes. Increased levels of C→T/G→A transitions were observed in old herbarium plastid DNA, representing 21.8% of observed miscoding lesions. We interpret this type of post-mortem DNA damage-derived modification to have arisen from the hydrolytic deamination of cytosine during long-term herbarium storage. Our results suggest that reliable sequence data can be obtained from herbarium specimens.     

How to Open the Treasure Chest? Optimising DNA Extraction from Herbarium Specimens

Herbarium collections are potentially an enormous resource for DNA studies, but the use of herbarium specimens in molecular studies has thus far been slowed down by difficulty in obtaining amplifiable DNA. Here we compare a set of commercially available DNA extraction protocols and their performance in terms of DNA purity and yield, and PCR amplification success as measured by using three differentially sized markers, the rbcL barcoding marker (cpDNA), the LEAFY exon 3 (nrDNA), and the trnL((UAA)) P6 loop (cpDNA). Results reveal large differences between extraction methods, where DNA purity rather than yield is shown to be strongly correlated with PCR success. Amplicon size shows similarly strong correlation with PCR success, with the shortest fragment showing the highest success rate (78%, P6 loop, 10-143 base pairs (bp)) and the largest fragment the lowest success (10%, rbcL, 670 bp). The effect of specimen preparation method on PCR success was also tested. Results show that drying method strongly affects PCR success, especially the availability of fragments longer than 250 bp, where longer fragments are more available for PCR amplification in air dried material compared to alcohol dried specimens. Results from our study indicate that projects relying on poor-quality starting material such as herbarium or scat samples should focus on extracting pure DNA and aim to amplify short target regions (<200-300 bp) in order to maximise outcomes. Development of shorter barcoding regions, or mini-barcodes within existing ones should be of high importance as only a few options are currently available; this is particularly important if we hope to incorporate the millions of herbarium samples available into barcoding initiatives and other molecular studies.     

中国高等植物数字化标本分析

传统上馆藏标本,主要用于植物分类学、植物资源学的研究。数字标本的出现将标本的使用拓展到从研究生物多样性时间空间分布到生态学和进化学理论、生物多样性保护、农业和人类健康等广泛领域。截至目前,从互联网上获取的采自中国的植物标本数量已有1 200多万份。该文通过整理和分析这些数据以了解中国植物标本的数字化精度、采集时间和采集地区规律以及采集空缺等状况。结果表明:中国标本采集形成了4个高峰,即20世纪30年代、60年代、80年代和21世纪初,中国植物标本采集和研究工作主要在20世纪50年代后由中国学者完成。标本采集地区覆盖度在省级较好,县级标本采集则很不平衡; 标本采集类群在科属层面覆盖率高,但近五分之一的物种采集不足; 标本的采集量既与植物分布幅度相关,也与采集地区的知名度、所获科研项目及采集者偏好有关。未来中国植物标本数字化方向应该在继续挖掘馆藏标本的同时,一方面开展对现有数字化标本信息再审核及补充,并加强与欧美大馆的信息共享以获取早期历史标本信息; 另一方面应用数字化标本信息分析结果,指导境内标本的精准采集,包括采集薄弱/空白地区、采集薄弱/空白属种的采集,以进一步增强实体标本馆能力,提高数字化标本质量,为进一步完善植物标本数字化和精准化采集提供依据,更好地服务科学和社会的发展。     

Increasing the efficiency of digitization workflows for herbarium specimens

The New York Botanical Garden Herbarium has been databasing and imaging its estimated 7.3 million plant specimens for the past 17 years. Due to the size of the collection, we have been selectively digitizing fundable subsets of specimens, making successive passes through the herbarium with each new grant. With this strategy, the average rate for databasing complete records has been 10 specimens per hour. With 1.3 million specimens databased, this effort has taken about 130,000 hours of staff time. At this rate, to complete the herbarium and digitize the remaining 6 million specimens, another 600,000 hours would be needed. Given the current biodiversity and economic crises, there is neither the time nor money to complete the collection at this rate.Through a combination of grants over the last few years, The New York Botanical Garden has been testing new protocols and tactics for increasing the rate of digitization through combinations of data collaboration, field book digitization, partial data entry and imaging, and optical character recognition (OCR) of specimen images. With the launch of the National Science Foundation's new Advancing Digitization of Biological Collections program, we hope to move forward with larger, more efficient digitization projects, capturing data from larger portions of the herbarium at a fraction of the cost and time.     

馆藏植物标本与iFlora

构建准确、有效的物种遗传信息库并运用其作为主要的物种鉴定检索数据库,是新一代植物志（iFlora）与传统植物志的本质区别。采集并准确鉴定数以万计的植物样品用于建立遗传信息库存在一定困难,随着遗传信息获取技术的快速发展,使馆藏植物标本成为获取遗传信息的重要补充,并具有提升遗传信息库建设速度和可靠性的较大潜能。本文结合已有的研究和实验室多年工作实践,论述了植物标本作为iFlora物种遗传信息提取材料的可行性和不可替代性;总结了标本遗传信息提取过程中的主要困难和存在问题,并提出一些解决方法;阐述了使用馆藏标本特别是利用模式标本材料构建遗传信息标准库对iFlora构建的重要作用和意义。     

Identifying dynamical persistent biomarker structures for rare events using modern integrative machine learning approach

The evolution of omics and computational competency has accelerated discoveries of the underlying biological processes in an unprecedented way. High throughput methodologies, such as flow cytometry, can reveal deeper insights into cell processes, thereby allowing opportunities for scientific discoveries related to health and diseases. However, working with cytometry data often imposes complex computational challenges due to high-dimensionality, large size, and nonlinearity of the data structure. In addition, cytometry data frequently exhibit diverse patterns across biomarkers and suffer from substantial class imbalances which can further complicate the problem. The existing methods of cytometry data analysis either predict cell population or perform feature selection. Through this study, we propose a “wisdom of the crowd” approach to simultaneously predict rare cell populations and perform feature selection by integrating a pool of modern machine learning (ML) algorithms. Given that our approach integrates superior performing ML models across different normalization techniques based on entropy and rank, our method can detect diverse patterns existing across the model features. Furthermore, the method identifies a dynamic biomarker structure that divides the features into persistently selected, unselected, and fluctuating assemblies indicating the role of each biomarker in rare cell prediction, which can subsequently aid in studies of disease progression.     

Biological collections in an ever changing world: Herbaria as tools for biogeographical and environmental studies

Plant specimens stored in herbaria are being used as never before to document the impacts of global change on humans and nature. However, published statistics on the use of biological collections are rare, and ecologists lack quantitative data demonstrating the relevance to science of herbarium specimens. I found 382 studies with original data that used herbarium specimens to document biogeographical patterns or environmental changes. Most studies are less than 10 years old, and only 1.4% of the herbarium specimens worldwide have been used to answer biogeographical or environmental questions. The vast majority (82%) of papers dealt with vascular plants, but some studies also used bryophytes, lichens, seaweeds and fungi. The herbarium specimens were collected from all continents, but most of the studies used specimens from North America (40% of studies) or Europe (28%). Many types of researches (conservation, plant disease, plant invasion, pollution, etc.) can be conducted using herbarium specimens. Climate change, and especially phenological reconstructions, are clearly emerging research topics. By group, small herbaria (<100,000 specimens) are consulted as often as very large herbaria (>1,000,000 specimens) for biogeographical and environmental research, but in most cases, only large facilities provide specimens collected worldwide. The median number of specimens per study in papers using computerized collections (15,295) was much higher than for papers that did not include electronic data (226). The use of molecular analyses to investigate herbarium specimens is still relatively unexplored, at least from biogeographical and environmental points of view. Combined with recently developed procedures to correct biases, herbarium specimens might provide in the near future exciting additional spatio-temporal insights that are currently unimaginable.     

使用Biotracks 采集植物标本

Biotracks 是一款自然观察类的公众科学应用,目前已经被各类科学调查和自然观察项目广泛使用。该文利用Biotracks 的标本采集项目将野外采集的数据与标本馆的数字馆藏系统连接起来,使用户在手机上记录的信息可以被应用到标本馆的标本数字化中。这种方式不仅提升了数字标本的转录效率,而且从根本上改变了整个标本收集流程中的数据整合方式,使得标本从采集到收藏的各个环节都能获得高质量的效率提升。同时,新的标本收集模式还能自然地将标本的野外照片与数字标本融为一体,从而使得传统标本原本很难呈现的颜色、行为、立体结构、环境等信息最终可以通过数字标本再次展现给研究者。这在信息维度上不仅拓展了传统标本的内涵,结合公众科学,未来还有望进一步延伸馆藏标本鉴定和讨论的时空范围。此外,公众科学在解决标本馆问题中所展现出来的潜质,为重新审视标本馆的领域价值提供了新的视角。     

Big Data to Knowledge: Application of Machine Learning to Predictive Modeling of Therapeutic Response in Cancer

BackgroundIn recent years, the availability of high throughput technologies, establishment of large molecular patient data repositories, and advancement in computing power and storage have allowed elucidation of complex mechanisms implicated in therapeutic response in cancer patients. The breadth and depth of such data, alongside experimental noise and missing values, requires a sophisticated human-machine interaction that would allow effective learning from complex data and accurate forecasting of future outcomes, ideally embedded in the core of machine learning design.ObjectiveIn this review, we will discuss machine learning techniques utilized for modeling of treatment response in cancer, including Random Forests, support vector machines, neural networks, and linear and logistic regression. We will overview their mathematical foundations and discuss their limitations and alternative approaches in light of their application to therapeutic response modeling in cancer.ConclusionWe hypothesize that the increase in the number of patient profiles and potential temporal monitoring of patient data will define even more complex techniques, such as deep learning and causal analysis, as central players in therapeutic response modeling.     

Metagenomic analysis of historical herbarium specimens reveals a postmortem microbial community

Advances in DNA extraction and next‐generation sequencing have made a vast number of historical herbarium specimens available for genomic investigation. These specimens contain not only genomic information from the individual plants themselves, but also from associated microorganisms such as bacteria and fungi. These microorganisms may have colonized the living plant (e.g., pathogens or host‐associated commensal taxa) or may result from postmortem colonization that may include decomposition processes or contamination during sample handling. Here we characterize the metagenomic profile from shotgun sequencing data from herbarium specimens of two widespread plant species (Ambrosia artemisiifolia and Arabidopsis thaliana) collected up to 180 years ago. We used blast searching in combination with megan and were able to infer the metagenomic community even from the oldest herbarium sample. Through comparison with contemporary plant collections, we identify three microbial species that are nearly exclusive to herbarium specimens, including the fungus Alternaria alternata, which can comprise up to 7% of the total sequencing reads. This species probably colonizes the herbarium specimens during preparation for mounting or during storage. By removing the probable contaminating taxa, we observe a temporal shift in the metagenomic composition of the invasive weed Am. artemisiifolia. Our findings demonstrate that it is generally possible to use herbarium specimens for metagenomic analyses, but that the results should be treated with caution, as some of the identified species may be herbarium contaminants rather than representing the natural metagenomic community of the host plant.     

Rethinking the links between systematic studies and ex situ living collections as a contribution to the Global Strategy for Plant Conservation

The Muséum national d’histoire naturelle (MNHN) in Paris holds ca. 70 million specimens. The collections were in need of a strategy to ensure their long-term conservation. We discuss how the Department of Botanical and Zoological Gardens (DJBZ; tropical living collections), and the Department of Systematics and Evolution (DSE; herbarium) contribute to achieving GSPC’s Target 1 (‘a widely accessible working list of known plant species as a step towards a complete world flora’). The DJBZ started encouraging better management of the collections, evolving towards focused reference collections, where all specimens have well-documented collection data. The objective is to link all collections to a scientific referee. This has already been achieved for a number of taxa. The herbarium of the DSE (acronym P) is among the world’s largest (11 million specimens, including 400,000 types). The collection’s heterogeneity impedes access to its data, since P is a mix of recent well-documented collections and historical collections at various curational levels. P is currently under renovation, which started by mounting all ca. 2 million unmounted specimens. The project also includes databasing and imaging of every specimen. The database now holds around 1,000,000 records. For taxonomic studies, living collections are crucial, especially for plants that are not easily preserved as herbarium specimens. Living collections also enable studies impossible to forecast at the time of collecting. Herbaria and living collections should therefore be conceived as interoperable entities requiring common scientific curation. Through a combination of its assets and the expertise of its researchers, the MNHN is well prepared to tackle the new objectives of the GSPC beyond 2010.     

Historical herbarium specimens in plant molecular systematics — an example from the fern genus <Emphasis Type="Italic">Lindsaea</Emphasis> (Lindsaeaceae)

We extracted, amplified and sequenced DNA from historical herbarium specimens and silica-dried samples of the fern genus Lindsaea in order to study the sequencing success between the two kinds of samples. High quality sequences were obtained from 57% of the herbarium samples. The specimens age was found to be of little importance for sequencing success when less than 75 years, but the colour of a specimen was found more indicative of sequencing success. Shorter DNA fragments were sequenced successfully twice as often as longer fragments from the herbarium material; in relatively recently collected silica-dried material longer sequences were obtained almost as frequently as short ones. No obvious differences in sequencing success between material originating from different herbaria was observed. We conclude that by using specifically designed DNA extraction protocols and by sequencing short DNA fragments from carefully selected specimens, herbarium material and type specimens can be successfully used in molecular systematics. Typical material or specimens from the type locality (topotypes) should be preferred, when placing a species in a phylogeny.     

Leaving flatland: Advances in 3D behavioral measurement

Animals move in three dimensions (3D). Thus, 3D measurement is necessary to report the true kinematics of animal movement. Existing 3D measurement techniques draw on specialized hardware, such as motion capture or depth cameras, as well as deep multi-view and monocular computer vision. Continued advances at the intersection of deep learning and computer vision will facilitate 3D tracking across more anatomical features, with less training data, in additional species, and within more natural, occlusive environments. 3D behavioral measurement enables unique applications in phenotyping, investigating the neural basis of behavior, and designing artificial agents capable of imitating animal behavior.