A bad tumor biomarker is as bad as a bad drug: The gap between genomics data and phenotype to predict response

Biological insights and novel biomarker discovery through deep learning approaches in breast cancer histopathology

Breast cancer remains a highly prevalent disease with considerable inter- and intra-tumoral heterogeneity complicating prognostication and treatment decisions. The utilization and depth of genomic, transcriptomic and proteomic data for cancer has exploded over recent times and the addition of spatial context to this information, by understanding the correlating morphologic and spatial patterns of cells in tissue samples, has created an exciting frontier of research, histo-genomics. At the same time, deep learning (DL), a class of machine learning algorithms employing artificial neural networks, has rapidly progressed in the last decade with a confluence of technical developments - including the advent of modern graphic processing units (GPU), allowing efficient implementation of increasingly complex architectures at scale; advances in the theoretical and practical design of network architectures; and access to larger datasets for training - all leading to sweeping advances in image classification and object detection. In this review, we examine recent developments in the application of DL in breast cancer histology with particular emphasis of those producing biologic insights or novel biomarkers, spanning the extraction of genomic information to the use of stroma to predict cancer recurrence, with the aim of suggesting avenues for further advancing this exciting field.

Tailoring therapies--improving the management of early breast cancer: St Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2015

The 14th St Gallen International Breast Cancer Conference (2015) reviewed substantial new evidence on locoregional and systemic therapies for early breast cancer. Further experience has supported the adequacy of tumor margins defined as 'no ink on invasive tumor or DCIS' and the safety of omitting axillary dissection in specific cohorts. Radiotherapy trials support irradiation of regional nodes in node-positive disease. Considering subdivisions within luminal disease, the Panel was more concerned with indications for the use of specific therapies, rather than surrogate identification of intrinsic subtypes as measured by multiparameter molecular tests. For the treatment of HER2-positive disease in patients with node-negative cancers up to 1 cm, the Panel endorsed a simplified regimen comprising paclitaxel and trastuzumab without anthracycline as adjuvant therapy. For premenopausal patients with endocrine responsive disease, the Panel endorsed the role of ovarian function suppression with either tamoxifen or exemestane for patients at higher risk. The Panel noted the value of an LHRH agonist given during chemotherapy for premenopausal women with ER-negative disease in protecting against premature ovarian failure and preserving fertility. The Panel noted increasing evidence for the prognostic value of commonly used multiparameter molecular markers, some of which also carried prognostic information for late relapse. The Panel noted that the results of such tests, where available, were frequently used to assist decisions about the inclusion of cytotoxic chemotherapy in the treatment of patients with luminal disease, but noted that threshold values had not been established for this purpose for any of these tests. Multiparameter molecular assays are expensive and therefore unavailable in much of the world. The majority of new breast cancer cases and breast cancer deaths now occur in less developed regions of the world. In these areas, less expensive pathology tests may provide valuable information. The Panel recommendations on treatment are not intended to apply to all patients, but rather to establish norms appropriate for the majority. Again, economic considerations may require that less expensive and only marginally less effective therapies may be necessary in less resourced areas. Panel recommendations do not imply unanimous agreement among Panel members. Indeed, very few of the 200 questions received 100% agreement from the Panel. In the text below, wording is intended to convey the strength of Panel support for each recommendation, while details of Panel voting on each question are available in supplementary Appendix S2, available at Annals of Oncology online.