Abstract
<title>Abstract</title> <p> HER2-targeted therapies require precise tumor diagnosis due to variability in HER2 expression. This study introduces a fully automated approach for analyzing fluorescent <italic>in situ</italic> hybridization (FISH) to determine HER2 status and resolve inconsistencies in immunohistochemistry (IHC) assessments. The <italic> <bold>g3mclass</bold> </italic> software was employed to identify HER2-low cases among HER2-negative breast cancers. Rather than imposing a predetermined number of FISH result groups, the software generated probability models using FISH data from 515 invasive cancers with diverse IHC scores and 52 non-cancerous tissues. The software established a normal HER2 copy number threshold of 2.11 and assigned each sample to the most probable group. Two distinct HER2-negative groups were identified. The HER2-normal group (217 cases) exhibited HER2 levels below the amplification threshold, consistent with normal gene activity. The HER2-low group (125 cases) demonstrated HER2 levels above the normal range ( <italic>P</italic> < 0.0001) but below the amplification threshold. IHC scores did not distinguish between HER2-low and HER2-normal groups, although the HER2-low group contained fewer IHC 0 cases (82.4% vs. 95.9%) and more IHC 1+ cases (15.2% vs. 3.2%). This software-assisted FISH methodology facilitates the identification of HER2-low tumors that may benefit from novel therapies and excludes those unlikely to respond, thereby supporting personalized treatment strategies. </p>