Gel electrophoresis is a ubiquitous laboratory method for the separation and semi-quantitative analysis of biomolecules. However, gel image analysis principles have barely advanced for decades, in stark contrast to other fields where AI has revolutionised data processing. Here, we show that an AI-based system can automatically identify gel bands in seconds for a wide range of experimental conditions, surpassing the capabilities of current software in both ease-of-use and versatility. We use a dataset containing 500+ images of manually-labelled gels to train various U-Nets to accurately identify bands through segmentation, i.e. classifying pixels as ‘band’ or ‘background’. When applied to gel electrophoresis data from other laboratories, our system generates results that quantitatively match those of the original authors. We have publicly released our models through GelGenie, an open-source application that allows users to extract bands from gel images on their own devices, with no expert knowledge or experience required.
18 March 2022 in IEEE Reviews in Biomedical Engineering
Technologies for quantifying circulating tumour DNA (ctDNA) in liquid biopsies could enable real-time measurements of cancer progression, profoundly impacting patient care. Sequencing methods can be too complex and time-consuming for regular point-of-care monitoring, but nanotechnology offers an alternative, harnessing the unique properties of objects tens to hundreds of nanometres in size. This systematic review was performed to identify all examples of nanotechnology-based ctDNA detection and assess their potential for clinical use. Google Scholar, PubMed, Web of Science, Google Patents, Espacenet and Embase/MEDLINE were searched up to 23rd March 2021. The review identified nanotechnology-based methods for ctDNA detection for which quantitative measures (e.g. limit of detection, LOD) were reported and biologically relevant samples were used. The pre-defined inclusion criteria were met by 66 records. LODs ranged from 10 zM to 50nM. 25 records presented an LOD of 10fM or below. Nanotechnology-based approaches could provide the basis for the next wave of advances in ctDNA diagnostics, enabling analysis at the point-of-care, but none are currently used clinically. Further work is needed in development and validation; trade-offs are expected between different performance measures e.g. number of sequences detected and time to result.
29 September 2021 in IEEE Signal Processing Letters
Convolutional Neural Networks (CNNs) have achieved impressive results across many super-resolution (SR) and image restoration tasks. While many such networks can upscale low-resolution (LR) images using just the raw pixel-level information, the ill-posed nature of SR can make it difficult to accurately super-resolve an image which has undergone multiple different degradations. Additional information (metadata) describing the degradation process (such as the blur kernel applied, compression level, etc.) can guide networks to super-resolve LR images with higher fidelity to the original source. Previous attempts at informing SR networks with degradation parameters have indeed been able to improve performance in a number of scenarios. However, due to the fully-convolutional nature of many SR networks, most of these metadata fusion methods either require a complete architectural change, or necessitate the addition of significant extra complexity. Thus, these approaches are difficult to introduce into arbitrary SR networks without considerable design alterations. In this paper, we introduce meta-attention, a simple mechanism which allows any SR CNN to exploit the information available in relevant degradation parameters. The mechanism functions by translating the metadata into a channel attention vector, which in turn selectively modulates the network's feature maps. Incorporating meta-attention into SR networks is straightforward, as it requires no specific type of architecture to function correctly. Extensive testing has shown that meta-attention can consistently improve the pixel-level accuracy of state-of-the-art (SOTA) networks when provided with relevant degradation metadata. Despite average memory/runtime overheads of less than ~2.6%/0.025 seconds for the datasets and models considered, meta-attention improves the performance for both PSNR and SSIM; for PSNR, the gain on blurred/downsampled (X4) images is of 0.2969 dB (on average) and 0.3320 dB for SOTA general and face SR models, respectively. The coding framework used for this paper is available at: https://github.com/um-dsrg/Super-Resolution-Meta-Attention-Networks.
Simultaneously studying different biomarker types (DNA, RNA, proteins, etc.) could improve understanding and diagnosis of many complex diseases. However, biomarker detection involves several complex or expensive methodologies, requiring specialized laboratories and personnel. A multiplexed assay would greatly facilitate the use of biomarker data. Here, we present a multiplexed biomarker detection technique using variable‐length DNA payload chains, which are systematically disassembled in the presence of specific biomarkers. The resulting distinctly‐sized fragments yield characteristic gel electrophoresis band patterns. This has enabled us to detect with high sensitivity and specificity DNA sequences including BRCA1 (limit of detection, LOD, ∼3 nM), RNA (miR‐141, LOD ∼19 nM) and the steroids aldosterone and cortisol (LOD ∼200–250 nM). We show that our assay is multiplexable, and suffers limited sensitivity loss in fetal bovine serum and can be applied using capillary electrophoresis, which may be more amenable to automation and integration in healthcare settings.
Crisscross polymerization enables the assembly of hundreds of unique DNA origami ‘slats’ into micron-sized structures with nanoscale precision. To design these megastructures, thousands of handle sequences from a fixed library must be assigned to individual slats to encode the desired binding architecture. This complexity presents two major challenges: handles must be selected to minimize parasitic interactions that compete with desired assembly, and the fabrication of hundreds of unique slats creates a substantial logistical burden. Here, we develop a unified framework that standardizes the design and fabrication of crisscross megastructures. We use an evolutionary algorithm to optimize handle assignment and minimize parasitic binding between slats. Together with an expanded handle library, the algorithm enables the assembly of large, multi-layered megastructures that otherwise would be produced at negligible yields. We have released this framework as #-CAD, an open-source graphical application that integrates these algorithms, streamlines laboratory workflows, and makes crisscross DNA origami more broadly accessible.