Weakly-supervised deep learning models enable HER2-low prediction from H &E stained slides.

BACKGROUND: Human epidermal growth factor receptor 2 (HER2)-low breast cancer has emerged as a new subtype of tumor, for which novel antibody-drug conjugates have shown beneficial effects. Assessment of HER2 requires several immunohistochemistry tests with an additional in situ hybridization test if a case is classified as HER2 2+. Therefore, novel cost-effective methods to speed up the HER2 assessment are highly desirable. METHODS: We used a self-supervised attention-based weakly supervised method to predict HER2-low directly from 1437 histopathological images from 1351 breast cancer patients. We built six distinct models to explore the ability of classifiers to distinguish between the HER2-negative, HER2-low, and HER2-high classes in different scenarios. The attention-based model was used to comprehend the decision-making process aimed at relevant tissue regions. RESULTS: Our results indicate that the effectiveness of classification models hinges on the consistency and dependability of assay-based tests for HER2, as the outcomes from these tests are utilized as the baseline truth for training our models. Through the use of explainable AI, we reveal histologic patterns associated with the HER2 subtypes. CONCLUSION: Our findings offer a demonstration of how deep learning technologies can be applied to identify HER2 subgroup statuses, potentially enriching the toolkit available for clinical decision-making in oncology.

A comparison between SOLiD 5500XLand Ion Torrent PGM-derived miRNA expression profiles in two breast cell lines.

Next-generation sequencing (NGS) platforms allow the analysis of hundreds of millions of molecules in a single sequencing run, revolutionizing many research areas. NGS-based microRNA studies enable expression quantification in unprecedented scale without the limitations of closed-platforms. Yet, whereas a massive amount of data produced by these platforms is available, comparisons of quantification/discovery capabilities between platforms are still lacking. Here we compare two NGS-platforms: SOLiD and PGM, by evaluating their microRNA identification/quantification capabilities using two breast-derived cell-lines. A high expression correlation (R2 > 0.9) was achieved, encompassing 97% of the miRNAs, and the few discrepancies in miRNA counts were attributable to molecules that have very low expression. Quantification divergences indicative of artefactual representation were seen for 14 miRNAs (higher in SOLiD-reads) and another 10 miRNAs more abundant in PGM-data. An inspection of these revealed an increased and statistically significant count of uracyls and uracyl-stretches for PGM-enriched miRNAs, compared to SOLiD and to the miRBase. In parallel, adenines and adenine-stretches were enriched for SOLiDderived miRNA reads. We conclude that, whereas both platforms are overall consistent and can be used interchangeably for microRNA expression studies, particular sequence features appear to be indicative of specific platform bias, and their presence in microRNAs should be considered for database-analyses.

A non-functional galanin receptor-2 in a multiple sclerosis patient.

Multiple Sclerosis (MS) is an inflammatory neurodegenerative disease that affects approximately 2.5 million people globally. Even though the etiology of MS remains unknown, it is accepted that it involves a combination of genetic alterations and environmental factors. Here, after performing whole exome sequencing, we found a MS patient harboring a rare and homozygous single nucleotide variant (SNV; rs61745847) of the G-protein coupled receptor (GPCR) galanin-receptor 2 (GALR2) that alters an important amino acid in the TM6 molecular toggle switch region (W249L). Nuclear magnetic resonance imaging showed that the hypothalamus (an area rich in GALR2) of this patient exhibited an important volumetric reduction leading to an enlarged third ventricle. Ex vivo experiments with patient-derived blood cells (AKT phosphorylation), as well as studies in recombinant cell lines expressing the human GALR2 (calcium mobilization and NFAT mediated gene transcription), showed that galanin (GAL) was unable to stimulate cell signaling in cells expressing the variant GALR2 allele. Live cell confocal microscopy showed that the GALR2 mutant receptor was primarily localized to intracellular endosomes. We conclude that the W249L SNV is likely to abrogate GAL-mediated signaling through GALR2 due to the spontaneous internalization of this receptor in this patient. Although this homozygous SNV was rare in our MS cohort (1:262 cases), our findings raise the potential importance of impaired neuroregenerative pathways in the pathogenesis of MS, warrant future studies into the relevance of the GAL/GALR2 axis in MS and further suggest the activation of GALR2 as a potential therapeutic route for this disease.

Machine learning predicts treatment sensitivity in multiple myeloma based on molecular and clinical information coupled with drug response.

Providing treatment sensitivity stratification at the time of cancer diagnosis allows better allocation of patients to alternative treatment options. Despite many clinical and biological risk markers having been associated with variable survival in cancer, assessing the interplay of these markers through Machine Learning (ML) algorithms still remains to be fully explored. Here, we present a Multi Learning Training approach (MuLT) combining supervised, unsupervised and self-supervised learning algorithms, to examine the predictive value of heterogeneous treatment outcomes for Multiple Myeloma (MM). We show that gene expression values improve the treatment sensitivity prediction and recapitulates genetic abnormalities detected by Fluorescence in situ hybridization (FISH) testing. MuLT performance was assessed by cross-validation experiments, in which it predicted treatment sensitivity with 68.70% of AUC. Finally, simulations showed numerical evidences that in average 17.07% of patients could get better response to a different treatment at the first line.

Deep Learning Predicts Underlying Features on Pathology Images with Therapeutic Relevance for Breast and Gastric Cancer.

DNA repair deficiency (DRD) is an important driver of carcinogenesis and an efficient target for anti-tumor therapies to improve patient survival. Thus, detection of DRD in tumors is paramount. Currently, determination of DRD in tumors is dependent on wet-lab assays. Here we describe an efficient machine learning algorithm which can predict DRD from histopathological images. The utility of this algorithm is demonstrated with data obtained from 1445 cancer patients. Our method performs rather well when trained on breast cancer specimens with homologous recombination deficiency (HRD), AUC (area under curve) = 0.80. Results for an independent breast cancer cohort achieved an AUC = 0.70. The utility of our method was further shown by considering the detection of mismatch repair deficiency (MMRD) in gastric cancer, yielding an AUC = 0.81. Our results demonstrate the capacity of our learning-base system as a low-cost tool for DRD detection.