Deep Learning Pipeline for Automated Cell Profiling from Cyclic Imaging.

Recent advances in microscopy allow scientists to generate vast amounts of biological data from a single biopsy sample. Cyclic fluorescence microscopy, in particular, enables multiple targets to be detected simultaneously. This, in turn, has deepened our understanding of tissue composition, cell-to-cell interactions, and cell signaling. Unfortunately, analysis of these datasets can be time-prohibitive due to the sheer volume of data. In this paper, we present CycloNET, a computational pipeline tailored for analyzing raw fluorescent images obtained through cyclic immunofluorescence. The automated pipeline pre-processes raw image files, quickly corrects for translation errors between imaging cycles, and leverages a pre-trained neural network to segment individual cells and generate single-cell molecular profiles. We applied CycloNET to a dataset of 22 human samples from head and neck squamous cell carcinoma patients and trained a neural network to segment immune cells. CycloNET efficiently processed a large-scale dataset (17 fields of view per cycle and 13 staining cycles per specimen) in 10 minutes, delivering insights at the single-cell resolution and facilitating the identification of rare immune cell clusters. We expect that this rapid pipeline will serve as a powerful tool to understand complex biological systems at the cellular level, with the potential to facilitate breakthroughs in areas such as developmental biology, disease pathology, and personalized medicine.

Rapid Serial Immunoprofiling of the Tumor Immune Microenvironment by Fine Needle Sampling.

PURPOSE: There is increasing effort to discover and integrate predictive and/or prognostic biomarkers into treatment algorithms. While tissue-based methods can reveal tumor-immune cell compositions at a single time point, we propose that single-cell sampling via fine needle aspiration (FNA) can facilitate serial assessment of the tumor immune microenvironment (TME) with a favorable risk-benefit profile. EXPERIMENTAL DESIGN: Primary antibodies directed against 20 murine and 25 human markers of interest were chemically modified via a custom linker-bio-orthogonal quencher (FAST) probe. A FAST-FNA cyclic imaging and analysis pipeline were developed to derive quantitative response scores. Single cells were harvested via FNA and characterized phenotypically and functionally both in preclinical and human samples using the newly developed FAST-FNA assay. RESULTS: FAST-FNA samples analyzed manually versus the newly developed deep learning-assisted pipeline gave highly concordant results. Subsequently, an agreement analysis showed that FAST and flow cytometry of surgically resected tumors were positively correlated with an R2 = 0.97 in preclinical samples and an R2 = 0.86 in human samples with the detection of the relevant tumor and immune biomarkers of interest. Finally, the feasibility of applying this minimally invasive approach to analyze the TME during immunotherapy was assessed in patients with cancer revealing local antitumor immune programs. CONCLUSIONS: The FAST-FNA is an innovative technology that combines bio-orthogonal chemistry coupled with a computational analysis pipeline for the comprehensive profiling of single cells obtained through FNA. This is the first demonstration that the complex and rapidly evolving TME during treatment can be accurately and serially measured by simple FNA.

CytoPAN-Portable cellular analyses for rapid point-of-care cancer diagnosis.

Rapid, automated, point-of-care cellular diagnosis of cancer remains difficult in remote settings due to lack of specialists and medical infrastructure. To address the need for same-day diagnosis, we developed an automated image cytometry system (CytoPAN) that allows rapid breast cancer diagnosis of scant cellular specimens obtained by fine needle aspiration (FNA) of palpable mass lesions. The system is devoid of moving parts for stable operations, harnesses optimized antibody kits for multiplexed analysis, and offers a user-friendly interface with automated analysis for rapid diagnoses. Through extensive optimization and validation using cell lines and mouse models, we established breast cancer diagnosis and receptor subtyping in 1 hour using as few as 50 harvested cells. In a prospective patient cohort study (n = 68), we showed that the diagnostic accuracy was 100% for cancer detection and the receptor subtyping accuracy was 96% for human epidermal growth factor receptor 2 and 93% for hormonal receptors (ER/PR), two key biomarkers associated with breast cancer. A combination of FNA and CytoPAN offers faster, less invasive cancer diagnoses than the current standard (core biopsy and histopathology). This approach should enable the ability to more rapidly diagnose breast cancer in global and remote settings.

Point-of-care cervical cancer screening using deep learning-based microholography.

Most deaths (80%) from cervical cancer occur in regions lacking adequate screening infrastructures or ready access to them. In contrast, most developed countries now embrace human papillomavirus (HPV) analyses as standalone screening; this transition threatens to further widen the resource gap. Methods: We describe the development of a DNA-focused digital microholography platform for point-of-care HPV screening, with automated readouts driven by customized deep-learning algorithms. In the presence of high-risk HPV 16 or 18 DNA, microbeads were designed to bind the DNA targets and form microbead dimers. The resulting holographic signature of the microbeads was recorded and analyzed. Results: The HPV DNA assay showed excellent sensitivity (down to a single cell) and specificity (100% concordance) in detecting HPV 16 and 18 DNA from cell lines. Our deep learning approach was 120-folder faster than the traditional reconstruction method and completed the analysis in < 2 min using a single CPU. In a blinded clinical study using patient cervical brushings, we successfully benchmarked our platform's performance to an FDA-approved HPV assay. Conclusions: Reliable and decentralized HPV testing will facilitate cataloguing the high-risk HPV landscape in underserved populations, revealing HPV coverage gaps in existing vaccination strategies and informing future iterations.

Structural topology defines protective CD8+ T cell epitopes in the HIV proteome.

Mutationally constrained epitopes of variable pathogens represent promising targets for vaccine design but are not reliably identified by sequence conservation. In this study, we employed structure-based network analysis, which applies network theory to HIV protein structure data to quantitate the topological importance of individual amino acid residues. Mutation of residues at important network positions disproportionately impaired viral replication and occurred with high frequency in epitopes presented by protective human leukocyte antigen (HLA) class I alleles. Moreover, CD8+ T cell targeting of highly networked epitopes distinguished individuals who naturally control HIV, even in the absence of protective HLA alleles. This approach thereby provides a mechanistic basis for immune control and a means to identify CD8+ T cell epitopes of topological importance for rational immunogen design, including a T cell-based HIV vaccine.