Portal de Periódicos da CAPES

Barcodes Are a Useful Tool for Labeling and Tracking Ecological Samples

2014; Ecological Society of America; Volume: 95; Issue: 3 Linguagem: Inglês

10.1890/0012-9623-95.3.293

2327-6096

Adam J. Copp, Theodore A. Kennedy, Jeffrey D. Muehlbauer,

Research Data Management Practices

Barcodes are used to label and track just about everything these days. Look around your office, in your medicine cabinet, at the package you just received in the mail, or on the shelves of any shop in town, and you will immediately grasp the ubiquity of their use. Interestingly, railroads and supermarkets were the early pioneers of barcode development: the former needing a way to track railway car location and ownership on a national scale, the latter needing a way to track a diverse array of products and to decrease checkout times (Nelson 1997). Barcodes first came to use in the sciences via the field of medicine, and the medical literature contains hundreds of publications describing how this technology has reduced errors in patient specimen identification and handling, where error mitigation is crucial. In short, barcodes have been adopted by many industries, and in many fields they are now synonymous with asset tracking. In spite of their potential to efficiently organize "assets" (i.e., samples) and minimize human error, the use of barcodes has yet to gain widespread application in ecology. In an age where students take notes on laptops instead of paper, and where "text messaging" involves a smartphone rather than a ball-point pen, why do otherwise tech-savvy ecologists persist in hand-labeling samples? Why do we repeatedly transcribe long and unique identifiers at each step in the process of sample analysis, thereby wasting time and creating opportunities for transcription errors and data loss? Why are most sample storage areas only successfully navigable by the lab manager who personally shelved the samples? In the case of our large ecology lab—and, we suspect, in many others as well—the answer to these questions was perpetually, "bar-coding won't be worth the trouble." Recently, however, we realized this was no longer a sufficient answer when we started a new research project that involved collecting an additional thousands of samples each year; we decided to embrace the tangible benefits of an electronic labeling system, and implemented barcoding in our lab. To be clear, the use of barcoding in ecology is not completely novel, and there have been early adopters of this technology. For example, the Cedar Creek Long Term Ecological Research (LTER) site has been using barcodes since at least the mid-1990s to track the large number of samples collected in their long-term experimental grasslands (T. A. Kennedy, personal observation). Overall, however, Cedar Creek is an outlier: in informal e-mail surveys of LTER sites, only two of eight respondents used barcodes for sample identification or tracking, and even then their use was generally limited to certain samples or certain stages of sample analysis. Our objective in this article is to use our lab as a case study to highlight the potential of barcodes to simplify numerous aspects of sample collection and processing. Our lab group collects thousands of algae, invertebrate, and fish diet samples each year as a major component of our food web research on the Colorado River in Grand Canyon. At any time, several research projects are running simultaneously, each with its own sample types, lab processing procedures, and sample storage requirements. With this large and constant influx of samples to the lab, we began having trouble keeping track of the enormous sample queue that developed each field season. Further, and with the benefit of hindsight, we were investing an inordinate amount of time transcribing pencil and paper field notes and lengthy sample labels, in addition to the time wasted investigating mislabeled samples and deciphering illegible handwriting. We knew barcodes were being used by some ecologists to track samples and thought it could be a useful tool in our research, but we kept waiting for a specific motivation to spur implementation. This motivation finally came in the summer of 2012 when we launched a new research project that filled our sample storage area to capacity with more numerous and varied samples. We had samples of all sorts stuffed into various boxes and stacked haphazardly in every bit of available space in a large storage area, and technicians had no way of knowing where to look for samples of interest, except by going through a frustrating and laborious process of guessing and checking boxes for relevant sample sets. Clearly, it was time to somehow streamline this entire process of what retailers refer to as "asset management"; it was time to start using barcodes. The first step in our implementation of sample barcoding was to design a unique barcode identification and labeling scheme using readily available software (Box 1). We print these barcode labels in bulk with a dedicated label printer and affix them to sample containers (e.g., Petri dishes, polyethylene bottles of ethanol, or empty heat-seal bags) prior to fieldwork. Using commercially available asset-tracking software, we developed data entry fields and a storage location framework specific to each sample type. Barcoded sample containers are scanned into an asset database at the time of collection using handheld computers equipped with barcode scanners—the same gadgets you might see a clerk using at a retail store. Field technicians enter sample information using the handheld computer's touch-screen interface, utilizing a combination of dropdown menus and direct character entry. After field data collection is complete, handheld computers are synchronized with the central lab computer's asset database, which collates and stores sample collection information and the sample's location. Sample information, especially storage location, is easily updated at every stage thereafter throughout the lab processes (i.e., long-term storage, temporary storage, sample processing, archiving, and disposal) by using the handheld computer to scan the sample's barcode label and revise the associated data field. We use asset-tracking software that allows users to interface a central asset-tracking database with handheld computers equipped with barcode scanners (Box 1). In this way, asset data can be entered and updated remotely using barcode labels as asset identifiers. The asset-tracking database includes sorting, reporting, and export functions for asset data; individual assets (i.e., samples) can be created, updated, or deleted within the database. Prior to fieldwork, project leads can create their own unique data entry fields, which can be structured as text boxes, check boxes, dates, etc. to facilitate field data entry. Critically, our setup requires users to assign a storage location to each asset via a "Building" and "Room" hierarchy. This hierarchy is manipulated for specific storage situations (e.g., Building = Storage room X, Room = Cabinet Y, Shelf Z). Within the database, the barcode ID is stored in a data key field named "asset ID," which becomes associated with all relevant information related to user-defined data entry fields and storage location. In our laboratory setup, we use a proprietary software program to design, format, and print barcode labels (Box 1). This allows user-developed unique sample IDs to be converted into barcode format, which are then printed using a label printer. We print labels in which the unique ID (in human-recognizable text) is printed alongside the computer-scannable barcode, allowing users to identify an item's unique ID without the use of a barcode scanner if necessary. Barcodes can be printed on labels of different types (i.e., chemical resistant, waterproof) and sizes, and labels can be printed individually or in large batches. In the field, technicians input sample data using a handheld computer with an integrated barcode scanner (Box 1). The instruments used specifically in our lab are lightweight (~ 400 g), and ruggedized to resist dust and water and to withstand multiple drops of > 1 m in height and temperature extremes of -20° to 50 °C. We have found that the handheld computers can be used for a full day in the field on a single battery charge, and that these computers can be fully recharged in a few hours using a small generator or other field power source (e.g., an AC power inverter attached to a car battery). As with many large research projects, our field data collection often gets ahead of laboratory analysis, resulting in a backlog of samples. Thus, we first introduced barcodes into our workflow by retroactively barcoding and collating sample information and storage location associated with our large sample backlog. We reorganized our storage area and created a detailed storage location framework as part of this initial implementation. The location of high priority samples that would be processed first was subsequently identified using the asset-tracking database. These samples were then moved from long-term storage to a temporary storage area within the lab, and their location was updated using the handheld computer. As samples were processed, their location and disposition was updated again (e.g., archived or destroyed). As a result of our initial barcoding implementation, we experienced substantial benefits to lab productivity. First, it forced us to reorganize and assign a specific location to each sample. With such detailed storage location information, it was easy to locate priority samples utilizing features of the asset-tracking database. The portable, handheld computer gave us the freedom to locally update sample information and location as samples progressed throughout the lab. This recurrent update process was convenient and quick. Additionally, when samples were processed in the lab, the barcode identifier was singularly transcribed, thereby reducing note-taking burden, eliminating transcription errors, and allowing laboratory personnel to spend more time on sample analysis. Lastly, barcoding empowered the lab manager to quickly report on the status of individual samples and to provide precise, up-to-date estimates of overall sample burdens, project progress, and storage capacity to project managers. Although we quickly realized that using barcodes benefited our laboratory operations, we were slow to implement them for field data collection because our field site (the Colorado River in Grand Canyon) is notoriously hard on electronics. Nevertheless, in 2013 we took a laptop equipped with asset-tracking software, two handheld computers, and 2000 sample containers pre-labeled with barcodes into the field on an 18-day river trip. The experimental design called for 100+ sample collection containers (i.e., sticky traps used to capture winged adult stages of aquatic insects [Smith et al. 2014]) to be deployed in arrays at each of six sampling reaches throughout ~ 400 km within Grand Canyon. For each sample container, we needed to record basic collection information such as date, time, river reach, and multiple attributes associated with the container's position in the trap array. When we deployed the sample containers, we carried handheld computers equipped with asset-tracking software, with data entry templates developed for the requisite basic collection information. Two groups of people deployed sticky traps simultaneously, because our research design required that many traps be deployed under a time restriction. As each sample container was deployed, a technician scanned the associated barcode and entered the sample information into the handheld computer. If site-specific notes were required to describe trap or site peculiarities, they were entered into the handheld at the time of deployment or upon retrieval. After collection, samples were put into a storage container. After the sample containers and all final data were collected for each site, sample collection information from the handhelds was uploaded to the field laptop and proofed for errors. If handheld computers required charging, they were charged during evening downtime at camp. Backup copies of the asset-tracking database were saved and stored separately from the field laptop after completion of each site's data collation to provide redundancy against loss of any single electronic device. It became apparent early in this field campaign that barcoding would revolutionize our field sampling. Because hand-drawn labels and field notebooks were eliminated from the field data collection, there were no subsequent issues in deciphering handwriting. Recording collection information into the handheld took ~30 seconds per sample container; this was comparable to the amount of time that would have been required to just create and apply a complex hand-drawn label, not even including the time to record basic collection information into field notebooks and later transcribe the notes onto digital spreadsheets upon return to the lab. As a result, lag time between data collection and digital collation was more or less eliminated. Ad hoc changes to sample collection attributes did not hinder field data collection, as data entry templates were easy to update and modify. Checking the collection data for errors was easily done within the asset database in the evening, such that errors could be corrected almost in real time while the field sampling was still fresh in everyone's memory, rather than days or months later back in the laboratory. Errors generally included entering the wrong date or selecting the wrong sample attribute from a dropdown menu. As we gained confidence with using barcodes, we decided to retroactively implement barcode labeling in a citizen scientist-driven study of insect emergence throughout the Colorado River in Grand Canyon. For this study, we worked with 8 professional river guides to collect light trap samples at camp each night during 7–14 day commercial river trips. At the onset of the study in 2012, guides filled out datasheets with sample collection information and then created hand-drawn labels in duplicate, which they placed within the sample containers and also affixed to the outside of the container. River guides collected the samples in the evening after long workdays, and we discovered that mislabeling was not uncommon. Further, because samples were transported long distances in tightly packed containers that were subjected to extreme heat, large rapids, and rough handling, many external labels became worn and unreadable by the time they reached the lab for processing. No data were lost because each sample also contained an internal label, but internal labels had to be removed and read for samples to be cataloged and stored, which added time and an additional source of error to laboratory sample processing. To mitigate these issues, in 2013 we began sending guides into the field with sets of pre-printed, duplicate barcode labels: one for labeling the sample container and the other for labeling the associated data sheet. When samples were collected, guides placed the duplicate labels onto the sample container and on the datasheet, and internal labels with detailed sample information were still utilized as a precaution against identification errors. River guides dropped off samples and datasheets to our lab after collection trips, and samples were stored and processed using the barcode label as their unique identifier. Again, our lab continued to benefit from further implementation of barcodes for sample labeling and tracking. By having the guides label sample containers and datasheets with barcodes in the field, mislabeling was nearly eliminated. Every one of the 500 barcode labels was readable after return from the field, and each sample was definitively matched to its associated datasheet via the duplicate label. Guides saved time in the field, as the barcode label they affixed to the datasheet and sample container replaced the relatively long, unique identifier they were previously responsible for hand-labeling on the outside of each sample container. As in our other implementations, the barcode identifier was singularly transcribed throughout lab processes, which saved technicians time, reduced transcription errors associated with traditional, lengthy unique identifiers, and made locating samples easy. Finally, as this sample set was frequently subsampled for more detailed analyses, there was considerable further value in the relatively short barcode identifier. Implementing a barcode labeling and tracking system has provided a host of benefits to our research lab, and in many cases the breadth of these benefits was not apparent prior to implementation. Error mitigation and time savings over the long term are hard to quantify, though these goals are given the highest priority within our lab. In order to implement this barcoding system, our lab had to research, purchase, and learn how to use the software and equipment, design implementation strategies, troubleshoot inevitable technical issues, and continually modify various components of the system when more efficient or practical designs became apparent. Thus, the full advantages of the barcoding system were realized only after these "growing pains" were worked out and newer research projects were developed explicitly to leverage this technology. Large monitoring and research groups may stand to benefit the most from a barcoding system, because the generally well-structured sample collection and processing procedures associated with such efforts are ideal for asset tracking on a large scale. We collect and subsequently manage thousands of samples per year, a factor that was central to our decision to implement barcoding. Smaller research experiments or pilot studies, where data collection is variable, sample numbers are low, and processing procedures have yet to be routinized, are less likely to realize tangible benefits. As with any new technology purchase, we spent a lot of time researching hardware and software vendors and products. Based on our experience, the asset-tracking software was the most important purchasing decision. There are a variety of asset-tracking programs available commercially, though none are designed specifically for ecological research. We ultimately purchased the software that seemed to best fit our needs at the time (Box 1), but with the benefit of hindsight we recognize that other software may have functioned as well or better for some of our applications. Thus, we recommend each lab take into account its own unique circumstances and budget when deciding upon an asset-tracking software package. It is also worth noting that most, but not all, asset-tracking software requires a service contract for professional troubleshooting and updates, which may factor into the decision process for many labs on tight budgets. Similarly, some asset-tracking software includes appealing barcode label design features, but we chose to purchase separate software for label design in our lab. There are also quite a few options when it comes to handheld computers, with a range of pricing and functionality depending on your budget and specific application. The same is true for label printers: we chose a printer based on label size and printing method. Many ecologists may also benefit from choosing a thermal transfer printer, as we did, because this printing method is more durable and chemical resistant than other printing methods. We also kept field conditions in mind when we purchased barcode labels, and ultimately settled on 2 × 1 inch gloss polyester labels with acrylic adhesive because this type of label is scratch, chemical, and UV resistant, and hopefully as "field-proof" as possible. Our goal in this paper has been to promote the use of barcode technology in ecology, a discipline that has seemingly avoided (either by tradition or by unfamiliarity) this technological advancement. Certainly, our use of barcodes is not pioneering, because barcodes have been around for decades and are already widely used in medical labs and in some progressive ecology labs. Based on our extremely positive experiences with using barcodes, however, we believe that far more ecology labs would have adopted this technology if they had a basic framework and specific examples of successful implementation, which we have attempted to provide here. At this point the equipment and software are relatively inexpensive, rugged, and easy to use, and result in almost immediate benefits to sample collection, processing, storage efficiency, and the reduction of transcription errors. So, prior to your next field campaign, hold off on that big purchase of label tape and permanent markers, and get a quote for barcode hardware and software instead. You'll be glad you did. We thank Garrett Thibedeau for his expertise in handling the inevitable technical issues faced while implementing this system. We also thank Ed Marut for software and hardware purchasing support. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.