More Like this

1. Integrative biology of injury in animals

   https://doi.org/10.1111/brv.12894  

   Rennolds, Corey W. ; Bely, Alexandra E. ( September 2022 , Biological Reviews)

   Mechanical injury is a prevalent challenge in the lives of animals with myriad potential consequences for organisms, including reduced fitness and death. Research on animal injury has focused on many aspects, including the frequency and severity of wounding in wild populations, the short‐ and long‐term consequences of injury at different biological scales, and the variation in the response to injury within or among individuals, species, ontogenies, and environmental contexts. However, relevant research is scattered across diverse biological subdisciplines, and the study of the effects of injury has lacked synthesis and coherence. Furthermore, the depth of knowledge across injury biology is highly uneven in terms of scope and taxonomic coverage: much injury research is biomedical in focus, using mammalian model systems and investigating cellular and molecular processes, while research at organismal and higher scales, research that is explicitly comparative, and research on invertebrate and non‐mammalian vertebrate species is less common and often less well integrated into the core body of knowledge about injury. The current state of injury research presents an opportunity to unify conceptually work focusing on a range of relevant questions, to synthesize progress to date, and to identify fruitful avenues for future research. The central aim of this review is to synthesize research concerning the broad range of effects of mechanical injury in animals. We organize reviewed work by four broad and loosely defined levels of biological organization: molecular and cellular effects, physiological and organismal effects, behavioural effects, and ecological and evolutionary effects of injury. Throughout, we highlight the diversity of injury consequences within and among taxonomic groups while emphasizing the gaps in taxonomic coverage, causal understanding, and biological endpoints considered. We additionally discuss the importance of integrating knowledge within and across biological levels, including how initial, localized responses to injury can lead to long‐term consequences at the scale of the individual animal and beyond. We also suggest important avenues for future injury biology research, including distinguishing better between related yet distinct injury phenomena, expanding the subjects of injury research to include a greater variety of species, and testing how intrinsic and extrinsic conditions affect the scope and sensitivity of injury responses. It is our hope that this review will not only strengthen understanding of animal injury but will contribute to building a foundation for a more cohesive field of ‘injury biology’.

    

   more » « less
2. Ice matters: Life‐history strategies of two Antarctic seals dictate climate change eventualities in the Weddell Sea

   https://doi.org/10.1111/gcb.15828  

   Wege, Mia ; Salas, Leo ; LaRue, Michelle ( September 2021 , Global Change Biology)

   The impacts of climate change in Antarctica and the Southern Ocean are not uniform and ice‐obligate species with dissimilar life‐history characteristics will likely respond differently to their changing ecosystems. We use a unique data set of WeddellLeptonychotes weddelliiand crabeater seals' (CESs)Lobodon carcinophagabreeding season distribution in the Weddell Sea, determined from satellite imagery. We contrast the theoretical climate impacts on both ice‐obligate predators who differ in life‐history characteristics: CESs are highly specialized Antarctic krillEuphausia superbapredators and breed in the seasonal pack ice; Weddell seals (WESs) are generalist predators and breed on comparatively stable fast ice. We used presence–absence data and a suite of remotely sensed environmental variables to build habitat models. Each of the environmental predictors is multiplied by a ‘climate change score’ based on known responses to climate change to create a ‘change importance product’. Results show CESs are more sensitive to climate change than WESs. Crabeater seals prefer to breed close to krill, and the compounding effects of changing sea ice concentrations and sea surface temperatures, the proximity to krill and abundance of stable breeding ice, can influence their post‐breeding foraging success and ultimately their future breeding success. But in contrast to the Ross Sea, here WESs prefer to breed closer to larger colonies of emperor penguins (Aptenodytes forsteri). This suggests that the Weddell Sea may currently be prey‐abundant, allowing the only two air‐breathing Antarctic silverfish predators (Pleuragramma antarctica) (WESs and emperor penguins) to breed closer to each other. This is the first basin‐scale, region‐specific comparison of breeding season habitat in these two key Antarctic predators based on real‐world data to compare climate change responses. This work shows that broad‐brush, basin‐scale approaches to understanding species‐specific responses to climate change are not always appropriate, and regional models are needed—especially when designing marine protected areas.

    

   more » « less
3. Transforming the undergraduate curriculum – engaging first year students in authentic research experiences

   Cohen, R.E. ; Land, A.M. ; Martensen, B.F. ; Sharlin, D.S. ; and Smith, B.A. ( February 2021 , Integrative and comparative biology)

   The Department of Biological Sciences at Minnesota State University, Mankato has recently implemented a first year undergraduate research experience (called the Research Immersive Scholastic Experience in Biology program; RISEbio) designed to improve student success and engagement in biology. In this program, first year students exchange introductory biology labs for the RISEbio curriculum, where they learn basic laboratory, analytical and scientific reasoning skills before beginning authentic mentored research projects in their second and third semesters. Students in one of three research tracks examine the neural control of reproductive behavior by examining gene expression in the brain of the seasonally breeding green anole lizard (Anolis carolinensis). Working in groups, students gain experience with bioinformatics by examining preliminary RNA-seq data and selecting a gene of interest. Then, students design and test primers to amplify their gene of interest, followed by isolating RNA from the hypothalamus of breeding and non-breeding lizard brains. Lastly, students utilize their isolated RNA samples and validated primers in quantitative RT-PCR studies to determine if their gene of interest is differentially expressed in the anole brain. Preliminary work has identified melatonin receptor 1A (MTNR1A) as more highly expressed in the breeding compared to non-breeding anole hypothalamus, while corticotropin releasing hormone binding protein (CRHBP) expression does not differ seasonally. Together with other aspects of the RISEbio program, these early research experiences have led to increased student outcomes, including increased academic success and enhanced scientific motivation. 

   more » « less
4. Geographical associations with anthropogenic noise pollution for North American breeding birds

   https://doi.org/10.1111/geb.13016  

   Klingbeil, Brian T. ; La Sorte, Frank A. ; Lepczyk, Christopher A. ; Fink, Daniel ; Flather, Curtis H. ; Sheard, ed., Catherine ( October 2019 , Global Ecology and Biogeography)

   Anthropogenic noise pollution (ANP) is a globally invasive phenomenon impacting natural systems, but most research has occurred at local scales with few species. We investigated continental‐scale breeding season associations with ANP for 322 bird species to test whether small‐scale predictions related to breeding habitat, migratory behaviour, body mass and vocal traits are consistent at broad spatial extents for an extensive group of species.

   Conterminous USA.

   2004–2011.

   North American breeding birds.

   We calculated, for each species, the association between the breeding season and ANP, using spatially explicit estimates of ANP from the National Park Service and weekly estimates of probabilities of occurrence based on observations from the eBird citizen‐science database. We evaluated how the association of the breeding season for each species with ANP was related to expectations based on size, migratory behaviour and breeding habitat. For a subset of species, we used vocal trait data for song duration, pitch and complexity to evaluate hypotheses from the birdsong literature related to habitat complexity and sensitivity to ANP.

   Species that breed predominantly in anthropogenic environments were associated with twice the level of ANP (~7.4 dB) as species breeding in forested habitats (~3.2 dB). However, we did not find evidence to suggest that birds with higher‐pitched songs are more likely to be found in areas with higher levels of ANP. Residents and migratory species did not differ in associations with ANP, but songs were less complex among forest‐breeding species than non‐forest‐breeding species and increased in complexity with increasing ANP.

   Anthropogenic noise pollution is an important factor associated with breeding distributions of bird species in North America. Vocal traits could be useful to understand factors that affect sensitivity to ANP and to predict the potential impact of ANP, although future studies should aim to understand how and why patterns differ across spatial scales.

    

   more » « less
5. The Temple University Digital Pathology Corpus: The Breast Tissue Subset

   https://doi.org/10.1109/SPMB52430.2021.9672275  

   Wevodau, Z. ; Doshna, B. ; Jhala, N. ; Akhtar, I. ; Obeid, I. ; Picone, J. ( December 2021 , Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, I. (Ed.)

   The Neural Engineering Data Consortium (NEDC) is developing the Temple University Digital Pathology Corpus (TUDP), an open source database of high-resolution images from scanned pathology samples [1], as part of its National Science Foundation-funded Major Research Instrumentation grant titled “MRI: High Performance Digital Pathology Using Big Data and Machine Learning” [2]. The long-term goal of this project is to release one million images. We have currently scanned over 100,000 images and are in the process of annotating breast tissue data for our first official corpus release, v1.0.0. This release contains 3,505 annotated images of breast tissue including 74 patients with cancerous diagnoses (out of a total of 296 patients). In this poster, we will present an analysis of this corpus and discuss the challenges we have faced in efficiently producing high quality annotations of breast tissue. It is well known that state of the art algorithms in machine learning require vast amounts of data. Fields such as speech recognition [3], image recognition [4] and text processing [5] are able to deliver impressive performance with complex deep learning models because they have developed large corpora to support training of extremely high-dimensional models (e.g., billions of parameters). Other fields that do not have access to such data resources must rely on techniques in which existing models can be adapted to new datasets [6]. A preliminary version of this breast corpus release was tested in a pilot study using a baseline machine learning system, ResNet18 [7], that leverages several open-source Python tools. The pilot corpus was divided into three sets: train, development, and evaluation. Portions of these slides were manually annotated [1] using the nine labels in Table 1 [8] to identify five to ten examples of pathological features on each slide. Not every pathological feature is annotated, meaning excluded areas can include focuses particular to these labels that are not used for training. A summary of the number of patches within each label is given in Table 2. To maintain a balanced training set, 1,000 patches of each label were used to train the machine learning model. Throughout all sets, only annotated patches were involved in model development. The performance of this model in identifying all the patches in the evaluation set can be seen in the confusion matrix of classification accuracy in Table 3. The highest performing labels were background, 97% correct identification, and artifact, 76% correct identification. A correlation exists between labels with more than 6,000 development patches and accurate performance on the evaluation set. Additionally, these results indicated a need to further refine the annotation of invasive ductal carcinoma (“indc”), inflammation (“infl”), nonneoplastic features (“nneo”), normal (“norm”) and suspicious (“susp”). This pilot experiment motivated changes to the corpus that will be discussed in detail in this poster presentation. To increase the accuracy of the machine learning model, we modified how we addressed underperforming labels. One common source of error arose with how non-background labels were converted into patches. Large areas of background within other labels were isolated within a patch resulting in connective tissue misrepresenting a non-background label. In response, the annotation overlay margins were revised to exclude benign connective tissue in non-background labels. Corresponding patient reports and supporting immunohistochemical stains further guided annotation reviews. The microscopic diagnoses given by the primary pathologist in these reports detail the pathological findings within each tissue site, but not within each specific slide. The microscopic diagnoses informed revisions specifically targeting annotated regions classified as cancerous, ensuring that the labels “indc” and “dcis” were used only in situations where a micropathologist diagnosed it as such. Further differentiation of cancerous and precancerous labels, as well as the location of their focus on a slide, could be accomplished with supplemental immunohistochemically (IHC) stained slides. When distinguishing whether a focus is a nonneoplastic feature versus a cancerous growth, pathologists employ antigen targeting stains to the tissue in question to confirm the diagnosis. For example, a nonneoplastic feature of usual ductal hyperplasia will display diffuse staining for cytokeratin 5 (CK5) and no diffuse staining for estrogen receptor (ER), while a cancerous growth of ductal carcinoma in situ will have negative or focally positive staining for CK5 and diffuse staining for ER [9]. Many tissue samples contain cancerous and non-cancerous features with morphological overlaps that cause variability between annotators. The informative fields IHC slides provide could play an integral role in machine model pathology diagnostics. Following the revisions made on all the annotations, a second experiment was run using ResNet18. Compared to the pilot study, an increase of model prediction accuracy was seen for the labels indc, infl, nneo, norm, and null. This increase is correlated with an increase in annotated area and annotation accuracy. Model performance in identifying the suspicious label decreased by 25% due to the decrease of 57% in the total annotated area described by this label. A summary of the model performance is given in Table 4, which shows the new prediction accuracy and the absolute change in error rate compared to Table 3. The breast tissue subset we are developing includes 3,505 annotated breast pathology slides from 296 patients. The average size of a scanned SVS file is 363 MB. The annotations are stored in an XML format. A CSV version of the annotation file is also available which provides a flat, or simple, annotation that is easy for machine learning researchers to access and interface to their systems. Each patient is identified by an anonymized medical reference number. Within each patient’s directory, one or more sessions are identified, also anonymized to the first of the month in which the sample was taken. These sessions are broken into groupings of tissue taken on that date (in this case, breast tissue). A deidentified patient report stored as a flat text file is also available. Within these slides there are a total of 16,971 total annotated regions with an average of 4.84 annotations per slide. Among those annotations, 8,035 are non-cancerous (normal, background, null, and artifact,) 6,222 are carcinogenic signs (inflammation, nonneoplastic and suspicious,) and 2,714 are cancerous labels (ductal carcinoma in situ and invasive ductal carcinoma in situ.) The individual patients are split up into three sets: train, development, and evaluation. Of the 74 cancerous patients, 20 were allotted for both the development and evaluation sets, while the remain 34 were allotted for train. The remaining 222 patients were split up to preserve the overall distribution of labels within the corpus. This was done in hope of creating control sets for comparable studies. Overall, the development and evaluation sets each have 80 patients, while the training set has 136 patients. In a related component of this project, slides from the Fox Chase Cancer Center (FCCC) Biosample Repository (https://www.foxchase.org/research/facilities/genetic-research-facilities/biosample-repository -facility) are being digitized in addition to slides provided by Temple University Hospital. This data includes 18 different types of tissue including approximately 38.5% urinary tissue and 16.5% gynecological tissue. These slides and the metadata provided with them are already anonymized and include diagnoses in a spreadsheet with sample and patient ID. We plan to release over 13,000 unannotated slides from the FCCC Corpus simultaneously with v1.0.0 of TUDP. Details of this release will also be discussed in this poster. Few digitally annotated databases of pathology samples like TUDP exist due to the extensive data collection and processing required. The breast corpus subset should be released by November 2021. By December 2021 we should also release the unannotated FCCC data. We are currently annotating urinary tract data as well. We expect to release about 5,600 processed TUH slides in this subset. We have an additional 53,000 unprocessed TUH slides digitized. Corpora of this size will stimulate the development of a new generation of deep learning technology. In clinical settings where resources are limited, an assistive diagnoses model could support pathologists’ workload and even help prioritize suspected cancerous cases. ACKNOWLEDGMENTS This material is supported by the National Science Foundation under grants nos. CNS-1726188 and 1925494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. REFERENCES [1] N. Shawki et al., “The Temple University Digital Pathology Corpus,” in Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications, 1st ed., I. Obeid, I. Selesnick, and J. Picone, Eds. New York City, New York, USA: Springer, 2020, pp. 67 104. https://www.springer.com/gp/book/9783030368432. [2] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning.” Major Research Instrumentation (MRI), Division of Computer and Network Systems, Award No. 1726188, January 1, 2018 – December 31, 2021. https://www. isip.piconepress.com/projects/nsf_dpath/. [3] A. Gulati et al., “Conformer: Convolution-augmented Transformer for Speech Recognition,” in Proceedings of the Annual Conference of the International Speech Communication Association (INTERSPEECH), 2020, pp. 5036-5040. https://doi.org/10.21437/interspeech.2020-3015. [4] C.-J. Wu et al., “Machine Learning at Facebook: Understanding Inference at the Edge,” in Proceedings of the IEEE International Symposium on High Performance Computer Architecture (HPCA), 2019, pp. 331–344. https://ieeexplore.ieee.org/document/8675201. [5] I. Caswell and B. Liang, “Recent Advances in Google Translate,” Google AI Blog: The latest from Google Research, 2020. [Online]. Available: https://ai.googleblog.com/2020/06/recent-advances-in-google-translate.html. [Accessed: 01-Aug-2021]. [6] V. Khalkhali, N. Shawki, V. Shah, M. Golmohammadi, I. Obeid, and J. Picone, “Low Latency Real-Time Seizure Detection Using Transfer Deep Learning,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2021, pp. 1 7. https://www.isip. piconepress.com/publications/conference_proceedings/2021/ieee_spmb/eeg_transfer_learning/. [7] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning,” Philadelphia, Pennsylvania, USA, 2020. https://www.isip.piconepress.com/publications/reports/2020/nsf/mri_dpath/. [8] I. Hunt, S. Husain, J. Simons, I. Obeid, and J. Picone, “Recent Advances in the Temple University Digital Pathology Corpus,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2019, pp. 1–4. https://ieeexplore.ieee.org/document/9037859. [9] A. P. Martinez, C. Cohen, K. Z. Hanley, and X. (Bill) Li, “Estrogen Receptor and Cytokeratin 5 Are Reliable Markers to Separate Usual Ductal Hyperplasia From Atypical Ductal Hyperplasia and Low-Grade Ductal Carcinoma In Situ,” Arch. Pathol. Lab. Med., vol. 140, no. 7, pp. 686–689, Apr. 2016. https://doi.org/10.5858/arpa.2015-0238-OA. 

   more » « less