Biology subjects

Sowmya, A.

Publications and source records attributed to Sowmya, A..

4 recordsLinked to original sources

bioRxiv2022

Imbalanced classification for protein subcellularlocalisation with multilabel oversampling

MotivationSubcellular localisation of human proteins is essential to comprehend their functions and roles in physiological processes, which in turn helps in diagnostic and prognostic studies of pathological conditions and impacts clinical decision making. Since proteins reside at multiple locations at the same time and few subcellular locations host far more proteins than other locations, the computational task for their subcellular localisation is to train a multilabel classifier while handling data imbalance. In imbalanced data, minority classes are underrepresented, thus leading to a heavy bias towards the majority classes and the degradation of predictive capability for the minority classes. Furthermore, data imbalance in multilabel settings is an even more complex problem due to the coexistence of majority and minority classes. ResultsOur studies reveal that based on the extent of concurrence of majority and minority classes, oversampling of minority samples through appropriate data augmentation techniques holds promising scope for boosting the classification performance for the minority classes. We measured the magnitude of data imbalance per class and the concurrence of majority and minority classes in the dataset. Based on the obtained values, we identified minority and medium classes, and a new oversampling method is proposed that includes nonlinear mixup, geometric and colour transformations for data augmentation and a sampling approach to prepare minibatches. Performance evaluation on the Human Protein Atlas Kaggle challenge dataset shows that the proposed method is capable of achieving better predictions for minority classes than existing methods. AvailabilityData used in this study is available at https://www.kaggle.com/competitions/human-protein-atlas-image-classification/data. Contactyang.song1@unsw.edu.au Supplementary informationSupplementary data are available at Bioinformatics online.

bioinformatics↗

bioRxiv2022

Data augmentation for imbalanced blood cell image classification

Due to progression in cell-cycle or duration of storage, classification of morphological changes in human blood cells is important for correct and effective clinical decisions. Automated classification systems help avoid subjective outcomes and are more efficient. Deep learning and more specifically Convolutional Neural Networks have achieved state-of-the-art performance on various biomedical image classification problems. However, real-world data often suffers from the data imbalance problem, owing to which the trained classifier is biased towards the majority classes and does not perform well on the minority classes. This study presents an imbalanced blood cells classification method that utilises Wasserstein divergence GAN, mixup and novel nonlinear mixup for data augmentation to achieve oversampling of the minority classes. We also present a minority class focussed sampling strategy, which allows effective representation of minority class samples produced by all three data augmentation techniques and contributes to the classification performance. The method was evaluated on two publicly available datasets of immortalised human T-lymphocyte cells and Red Blood Cells. Classification performance evaluated using F1-score shows that our proposed approach outperforms existing methods on the same datasets.

bioinformatics↗

bioRxiv2022

Blood-based transcriptomic signature panel identification for cancer diagnosis: Benchmarking of feature extraction methods

Liquid biopsy has shown promise for cancer diagnosis due to its minimally invasive nature and the potential for novel biomarker discovery. However, the low concentration of relevant blood-based biosources and the heterogeneity of samples (i.e. the variability of relative abundance of molecules identified), pose major challenges to biomarker discovery. Moreover, the number of molecular measurements or features (e.g., transcript read counts) per sample could be in the order of several thousand, whereas the number of samples is often substantially lower, leading to the curse of dimensionality. These challenges, among others, elucidate the importance of a robust biomarker panel identification or feature extraction step wherein relevant molecular measurements are identified prior to classification for cancer detection. In this work, we performed a benchmarking study on eleven feature extraction methods using transcriptomic profiles derived from different blood-based biosources. The methods were assessed both in terms of their predictive performance and the robustness of the biomarker panels in diagnosing cancer or stratifying cancer subtypes. While performing the comparison, the feature extraction methods are categorised into feature subset selection methods and transformation methods. A transformation feature extraction method, namely PLS-DA, was found to perform consistently superior in terms of classification performance, while a feature subset selection method, namely Ranger, selected feature sets that were the most robust to sub-sampling in terms of consistently selecting the same features. As part of the benchmarking study, a generic pipeline has been created and made available as an R package to ensure reproducibility of the results and allow for easy extension of this study to other datasets.

bioinformatics↗

bioRxiv2020

Estimation of Three-Dimensional Chromatin Morphology for Nuclear Classification and Characterisation

Classification and characterisation of cellular morphological states are vital for understanding cell differentiation, development, proliferation and diverse pathological conditions. As the onset of morphological changes transpires following genetic alterations in the chromatin configuration inside the nucleus, the nuclear texture as one of the low-level properties if detected and quantified accurately has the potential to provide insights on nuclear organisation and enable early diagnosis and prognosis. This study presents a three dimensional (3D) nuclear texture description method for cell nucleus classification and variation measurement in chromatin patterns on the transition to another phenotypic state. The proposed approach includes third plane information using hyperplanes into the design of the Sorted Random Projections (SRP) texture feature. The significance of including third plane information for low-resolution volumetric images is investigated by comparing the performance of 3D texture descriptor with its respective pseudo 3D form that ignores the interslice intensity correlations. Following classification, changes in chromatin pattern are estimated by computing the ratio of heterochromatin and euchromatin corresponding to their respective intensities and image gradient obtained by 3D SRP. The proposed method is evaluated on two publicly available 3D image datasets of human fibroblast and human prostate cancer cell lines in two phenotypic states obtained from the public Statistics Online Computational Resource. Experimental results show that 3D SRP and 3D Local Binary Pattern provide better results than other utilised handcrafted descriptors and deep learning features extracted using a pre-trained model. The results also show the advantage of utilising 3D feature descriptor for classification over its corresponding pseudo version. In addition, the proposed method validates that as the cell passes to another phenotypic state, there is a change in intensity and aggregation of heterochromatin. Author SummaryAutomated classification and measurement of cellular phenotypic traits can significantly impact clinical decision making. Early detection of diseases requires an accurate description of low-level cellular features to detect small-scale abnormalities in the few abnormal cells in the tissue microenvironment. The challenge is the development of a computational approach for 3D textural feature description that can capture the heterogeneous information in multiple dimensions and characterise the cells in their ultimate and intermediate phenotypic states effectively. Our work has proposed the method and metrics to measure chromatin condensation pattern and classify the phenotypic states. Experimental evaluation on the 3D image set of human fibroblast and human prostate cancer cell collections validates the proposed method for the classification of cell states. Results also signify the credibility of proposed metrics to characterise the cellular phenotypic states and contributes to studies related to early diagnosis, prognosis and drug resistance.

bioinformatics↗