Publications

In mammography, calcifications are one of the most common signs of breast cancer. Detection of such lesions is an active area of research for computer-aided diagnosis and machine learning algorithms. Due to limited numbers of positive cases, many supervised detection models suffer from overfitting and fail to generalize. We present a one-class, semi-supervised framework using a deep convolutional autoencoder trained with over 50,000 images from 11,000 negative-only cases. Since the model learned from only normal breast parenchymal features, calcifications produced large signals when comparing the residuals between input and reconstruction output images. As a key advancement, a structural dissimilarity index was used to suppress non-structural noises. Our selected model achieved pixel-based AUROC of 0.959 and AUPRC of 0.676 during validation, where calcification masks were defined in a semi-automated process. Although not trained directly on any cancers, detection performance of calcification lesions on 1,883 testing images (645 malignant and 1238 negative) achieved 75% sensitivity at 2.5 false positives per image. Performance plateaued early when trained with only a fraction of the cases, and greater model complexity or a larger dataset did not improve performance. This study demonstrates the potential of this anomaly detection approach to detect mammographic calcifications in a semi-supervised manner with efficient use of a small number of labeled images, and may facilitate new clinical applications such as computer-aided triage and quality improvement.

The colorectal cancer (CRC)-associated microbiota creates a pro-tumorigenic intestinal milieu and shapes immune responses within the tumor microenvironment. However, how oncomicrobes – like Fusobacterium nucleatum, found in the oral cavity and associated with CRC tissues– affect these distinct aspects of tumorigenesis is difficult to parse. Herein, we found that neonatal inoculation of Apc^Min/+ mice with F. nucleatum strain Fn7-1 circumvents technical barriers preventing its intestinal colonization, drives colonic Il17a expression prior to tumor formation, and potentiates intestinal tumorigenesis. Using gnotobiotic mice colonized with a minimal complexity microbiota (the altered Schaedler’s flora), we observed that intestinal Fn7-1 colonization increases colonic Th17 cell frequency and their IL-17A and IL-17F expression, along with a concurrent increase in colonic lamina propria Il23p19 expression. As Fn7-1 stably colonizes the intestinal tract in our models, we posited that microbial metabolites, specifically short-chain fatty acids (SCFA) that F. nucleatum abundantly produces in culture and, as we demonstrate, in the intestinal tract, might mediate part of its immunomodulatory effects in vivo. Supporting this hypothesis, we found that Fn7-1 did not alter RORγt⁺ CD4⁺T cell frequency in the absence of the SCFA receptor FFAR2. Taken together, our work suggests that F. nucleatum influences intestinal immunity by shaping Th17 responses in an FFAR2-dependent manner, although further studies are necessary to clarify the precise and multifaceted roles of FFAR2. The potential to increase intestinal Th17 responses is shared by another oncomicrobe, enterotoxigenic Bacteroides fragilis, highlighting a conserved pathway that could potentially be targeted to slow oncomicrobe-mediated CRC.

A more complete and holistic view on host–microbe interactions is needed to understand the physiological and cellular barriers that affect the efficacy of drug treatments and allow the discovery and development of new therapeutics. Here, we developed a multimodal imaging approach combining histopathology with mass spectrometry imaging (MSI) and same section imaging mass cytometry (IMC) to study the effects of Salmonella Typhimurium infection in the liver of a mouse model using the S. Typhimurium strains SL3261 and SL1344. This approach enables correlation of tissue morphology and specific cell phenotypes with molecular images of tissue metabolism. IMC revealed a marked increase in immune cell markers and localization in immune aggregates in infected tissues. A correlative computational method (network analysis) was deployed to find metabolic features associated with infection and revealed metabolic clusters of acetyl carnitines, as well as phosphatidylcholine and phosphatidylethanolamine plasmalogen species, which could be associated with pro-inflammatory immune cell types. By developing an IMC marker for the detection of Salmonella LPS, we were further able to identify and characterize those cell types which contained S. Typhimurium.

Image resolution in mass spectrometry imaging (MSI) is governed by the sampling probe, the motion of the stage relative to the probe, and the noise inherent for the sample and instrumentation employed. A new image formation model accounting for these variables is presented here. The model shows that the size of the probe, stage velocity, and the rate at which the probe consumes material from the surface govern the amount of blur present in the image. However, the main limiting factor for resolution is the signal-to-noise ratio (SNR). To evaluate blurring and noise effects, a new computational method for measuring lateral resolution in MSI is proposed. A spectral decomposition of the observed image signal and noise is used to determine a resolution number. To evaluate this technique, a silver step edge was prepared. This device was imaged at different pixels sizes using matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI). A modulation transfer function (MTF) and a noise power spectrum (NPS) were computed for each single-ion image, and resolution was defined as the point of intersection between the MTF and the NPS. Finally, the algorithm was also applied to a MALDI MSI tissue data set.

Background: Mutational signatures proved to be a useful tool for identifying patterns of mutations in genomes, often providing valuable insights about mutagenic processes or normal DNA damage. De novo extraction of signatures is commonly performed using Non-Negative Matrix Factorisation methods, however, accurate attribution of these signatures to individual samples is a distinct problem requiring uncertainty estimation, particularly in noisy scenarios or when the acting signatures have similar shapes. Whilst many packages for signature attribution exist, a few provide accuracy measures, and most are not easily reproducible nor scalable in high-performance computing environments.

Results: We present Mutational Signature Attribution (MSA), a reproducible pipeline designed to assign signatures of diferent mutation types on a single-sample basis, using Non-Negative Least Squares method with optimisation based on confgurable simulations. Parametric bootstrap is proposed as a way to measure statistical uncertainties of signature attribution. Supported mutation types include single and doublet base substitutions, indels and structural variants. Results are validated using simulations with reference COSMIC signatures, as well as randomly generated signatures.

Conclusions: MSA is a tool for optimised mutational signature attribution based on simulations, providing confdence intervals using parametric bootstrap. It comprises a set of Python scripts unifed in a single Nextfow pipeline with containerisation for cross-platform reproducibility and scalability in high-performance computing environments. The tool is publicly available from https://gitlab.com/s.senkin/MSA.

Keywords: MSA, Mutational signatures, NNLS, Parametric bootstrap, Nextfow