RT Journal Article SR Electronic T1 Functional assessments of PTEN variants using machine-assisted phenotype scoring JF bioRxiv FD Cold Spring Harbor Laboratory SP 2020.10.16.342915 DO 10.1101/2020.10.16.342915 A1 Jesse T. Chao A1 Calvin D. Roskelley A1 Christopher J.R. Loewen YR 2020 UL http://biorxiv.org/content/early/2020/10/16/2020.10.16.342915.abstract AB Genetic testing is widely used in evaluating a patient’s predisposition for developing a malignancy. In the case of cancer, when a functionally impactful inherited mutation (i.e. genetic variant) is identified in a disease-relevant gene, the patient is at elevated risk of developing a lesion in their lifetime. Unfortunately, as the rate and coverage of genetic testing has accelerated, our ability to make informed assessments regarding the functional status of the variants has lagged. Currently, there are two main strategies for assessing variant functions: in silico predictions or in vitro testing. The first approach is to build generalist computational prediction software using theoretical parameters such as amino acid conservation as feature inputs. These types of software can classify any variant of any gene. Although versatile, their non-specific design and theoretical assumptions result in different models frequently producing conflicting classifications. The second approach is to develop gene-specific assays. Although each assay is tailored to the physiological function of the gene, this approach requires significant investments. Therefore, there is an urgent need for more practical, streamlined and cost-effective methods. To directly address these issues, we designed a new approach of using alterations in protein subcellular localization as a key indicator of loss of function. Thus, new variants can be rapidly functionalized by using high-content microscopy. To facilitate the analysis of large amounts of image data, we developed a new software, named MAPS for machine-assisted phenotype scoring, that utilizes deep learning (DL) techniques to extract and classify cell-level phenotypes. This new Python-based toolkit helps users leverage commercial cloud-based DL services that are easy to train and deploy to fit varying experimental conditions. Model training is entirely code-free and can be done with limited number of images. Users simply input the trained endpoints into MAPS to accomplish cell detection, phenotype discovery and phenotype classification. Thus, MAPS allows cell biologists to easily apply DL to accelerate their image analysis workflow.Competing Interest StatementThe authors have declared no competing interest.