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

Autor(es)

Valieris, Renan ^[1] ; Amaro, Lucas ^[1] ; Bueno de Toledo Osorio, Cynthia Aparecida ^[2] ; Bueno, Adriana Passos ^{[1, 2]} ; Rosales Mitrowsky, Rafael Andres ^[3] ; Carraro, Dirce Maria ^[4] ; Nunes, Diana Noronha ^[5] ; Dias-Neto, Emmanuel ^[5] ; da Silva, Israel Tojal ^[1]

Afiliação do(s) autor(es)

^[1] CIPE AC Camargo Canc Ctr, Lab Computat Biol Bioinformat, BR-01508010 Sao Paulo - Brazil

^[2] CIPE AC Camargo Canc Ctr, Dept Pathol, BR-01525001 Sao Paulo - Brazil

^[3] Univ Sao Paulo, Dept Computat & Math, BR-14040901 Ribeirao Preto - Brazil

^[4] CIPE AC Camargo Canc Ctr, Lab Genom & Mol Biol, BR-01508010 Sao Paulo - Brazil

^[5] CIPE AC Camargo Canc Ctr, Med Genom Lab, BR-01525001 Sao Paulo - Brazil

• Tipo de documento: Artigo Científico
• Fonte: CANCERS; v. 12, n. 12 DEC 2020.
• Citações Web of Science: 0

Resumo

Simple Summary DNA repair deficiency (DRD) is common in many cancers. This deficiency contributes to pathogenesis of the disease, but it also presents an opportunity for therapeutic targeting. However, current DRD identification assays are not available for all patients. We propose an efficient machine learning algorithm which can predict DRD from histopathological images. The utility of our method was shown by considering the detection of homologous recombination deficiency (HRD) and mismatch repair deficiency (MMRD) in breast and gastric cancer respectively. Our findings demonstrate that machine-learning approaches can be used in advanced applications to assist therapy decisions. 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. (AU)

• Processo FAPESP: 15/19324-6 - Retrotransposons e enzimas de edição de ácidos nucléicos: ativação, eventos somáticos e associação com o câncer
• Beneficiário: Israel Tojal da Silva
• Modalidade de apoio: Auxílio à Pesquisa - Regular
• Processo FAPESP: 14/26897-0 - Epidemiologia e genômica de adenocarcinomas gástricos no Brasil
• Beneficiário: Emmanuel Dias-Neto
• Modalidade de apoio: Auxílio à Pesquisa - Temático