3 February 2003 in Nature Genetics
12 August 2021 in Nature Communications
Pathology is practiced by visual inspection of histochemically stained tissue slides. While the hematoxylin and eosin (H&E) stain is most commonly used, special stains can provide additional contrast to different tissue components. Here, we demonstrate the utility of supervised learning-based computational stain transformation from H&E to special stains (Masson’s Trichrome, periodic acid-Schiff and Jones silver stain) using kidney needle core biopsy tissue sections. Based on the evaluation by three renal pathologists, followed by adjudication by a fourth pathologist, we show that the generation of virtual special stains from existing H&E images improves the diagnosis of several non-neoplastic kidney diseases, sampled from 58 unique subjects (P = 0.0095). A second study found that the quality of the computationally generated special stains was statistically equivalent to those which were histochemically stained. This stain-to-stain transformation framework can improve preliminary diagnoses when additional special stains are needed, also providing significant savings in time and cost.
2 November 2014 in Nature
Whole-exome sequencing is used to compare the mutational landscape of adenomas from three mouse models of non-small-cell lung cancer, induced either by exposure to carcinogens or by genetic mutation of Kras; the results reveal that the two types of tumour have different mutational profiles and adopt different routes to tumour development. Allan Balmain and colleagues use whole-exome sequencing to compare the mutational landscape of adenomas from three mouse models of non-small cell lung cancer, induced by exposure to the carcinogens methyl-nitrosourea (MNU) and urethane, or by genetic activation of Kras (KrasLA2). Although the MNU-induced and KrasLA2 tumours carried the same initiating Kras mutation, MNU tumours also carry numerous non-synonymous point mutations. At the same time, KrasLA2 tumours carried numerous copy number alterations. This suggests that carcinogen-induced and genetically engineered models adopt different routes to tumour development. The results argue for a major role of germline Kras status in mutation selection during initiation. Collectively, these data demonstrate the utility of carcinogen models for understanding the complex mutation spectra seen in human cancers. Next-generation sequencing of human tumours has refined our understanding of the mutational processes operative in cancer initiation and progression, yet major questions remain regarding the factors that induce driver mutations and the processes that shape mutation selection during tumorigenesis. Here we performed whole-exome sequencing on adenomas from three mouse models of non-small-cell lung cancer, which were induced either by exposure to carcinogens (methyl-nitrosourea (MNU) and urethane) or by genetic activation of Kras (KrasLA2). Although the MNU-induced tumours carried exactly the same initiating mutation in Kras as seen in the KrasLA2 model (G12D), MNU tumours had an average of 192 non-synonymous, somatic single-nucleotide variants, compared with only six in tumours from the KrasLA2 model. By contrast, the KrasLA2 tumours exhibited a significantly higher level of aneuploidy and copy number alterations compared with the carcinogen-induced tumours, suggesting that carcinogen-induced and genetically engineered models lead to tumour development through different routes. The wild-type allele of Kras has been shown to act as a tumour suppressor in mouse models of non-small-cell lung cancer. We demonstrate that urethane-induced tumours from wild-type mice carry mostly (94%) Kras Q61R mutations, whereas those from Kras heterozygous animals carry mostly (92%) Kras Q61L mutations, indicating a major role for germline Kras status in mutation selection during initiation. The exome-wide mutation spectra in carcinogen-induced tumours overwhelmingly display signatures of the initiating carcinogen, while adenocarcinomas acquire additional C > T mutations at CpG sites. These data provide a basis for understanding results from human tumour genome sequencing, which has identified two broad categories of tumours based on the relative frequency of single-nucleotide variations and copy number alterations1, and underline the importance of carcinogen models for understanding the complex mutation spectra seen in human cancers.
5 September 2019 in Journal of the American Society of Nephrology
Significance Statement: Pathologists usually classify diabetic nephropathy on the basis of a visual assessment of glomerular pathology. Although diagnostic guidelines are well established, results may vary among pathologists. Modern machine learning has the potential to automate and augment accurate and precise classification of diabetic nephropathy. Digital algorithms may also be able to extract novel features relevant to disease progression and prognosis. The authors used image analysis and machine learning algorithms to digitally classify biopsy samples from 54 patients with diabetic nephropathy and found substantial agreement between digital classifications and those by three different pathologists. The study demonstrates that digital processing of renal tissue may provide useful information that may augment traditional clinical diagnostics. Background: Pathologists use visual classification of glomerular lesions to assess samples from patients with diabetic nephropathy (DN). The results may vary among pathologists. Digital algorithms may reduce this variability and provide more consistent image structure interpretation. Methods: We developed a digital pipeline to classify renal biopsies from patients with DN. We combined traditional image analysis with modern machine learning to efficiently capture important structures, minimize manual effort and supervision, and enforce biologic prior information onto our model. To computationally quantify glomerular structure despite its complexity, we simplified it to three components consisting of nuclei, capillary lumina and Bowman spaces; and Periodic Acid-Schiff positive structures. We detected glomerular boundaries and nuclei from whole slide images using convolutional neural networks, and the remaining glomerular structures using an unsupervised technique developed expressly for this purpose. We defined a set of digital features which quantify the structural progression of DN, and a recurrent network architecture which processes these features into a classification. Results: Our digital classification agreed with a senior pathologist whose classifications were used as ground truth with moderate Cohen’s kappa κ = 0.55 and 95% confidence interval [0.50, 0.60]. Two other renal pathologists agreed with the digital classification with κ1 = 0.68, 95% interval [0.50, 0.86] and κ2 = 0.48, 95% interval [0.32, 0.64]. Our results suggest computational approaches are comparable to human visual classification methods, and can offer improved precision in clinical decision workflows. We detected glomerular boundaries from whole slide images with 0.93±0.04 balanced accuracy, glomerular nuclei with 0.94 sensitivity and 0.93 specificity, and glomerular structural components with 0.95 sensitivity and 0.99 specificity. Conclusions: Computationally derived, histologic image features hold significant diagnostic information that may augment clinical diagnostics.
15 January 2014 in Cancer Research
The ubiquitin ligase CUL4A has been implicated in tumorigenesis, but its contributions to progression and metastasis have not been evaluated. Here, we show that CUL4A is elevated in breast cancer as well as in ovarian, gastric, and colorectal tumors in which its expression level correlates positively with distant metastasis. CUL4A overexpression in normal or malignant human mammary epithelial cells increased their neoplastic properties in vitro and in vivo, markedly increasing epithelial–mesenchymal transition (EMT) and the metastatic capacity of malignant cells. In contrast, silencing CUL4A in aggressive breast cancer cells inhibited these processes. Mechanistically, we found that CUL4A modulated histone H3K4me3 at the promoter of the EMT regulatory gene ZEB1 in a manner associated with its transcription. ZEB1 silencing blocked CUL4A-driven proliferation, EMT, tumorigenesis, and metastasis. Furthermore, in human breast cancers, ZEB1 expression correlated positively with CUL4A expression and distant metastasis. Taken together, our findings reveal a pivotal role of CUL4A in regulating the metastatic behavior of breast cancer cells. Cancer Res; 74(2); 520–31. ©2013 AACR.