Page 94 - Annual report 2021-22

Annual Report 2021-22 |

Vamsi Yenamandra

77

Vamsi Yenamandra is engaged in development of `Genoderm Matrix`, an online phenotype-genotype
characterization tool, to enable a rapid, scalable and cost-effective diagnostic approach in patients
with genodermatoses.

This tool aims to compute the most probable diagnosis for a genetic skin disease called epidermolysis
bullosa (EB) on the basis of severity, presence and absence of clinical features. Briefly, the tool consists
of a 19 x 9 size matrix, in which the 19 rows correspond to each of the clinical features and the 9 rows
correspond to each of the possible EB diagnoses. A mapping method is used to fill each cell of the
matrix with the possible values for each feature being -, +, ++ or not applicable (NA), according to the
observed clinical feature. These are then converted into binary digits 1 or 0 depending on whether it
is a static feature, absence feature, or a variable feature. The scores of each of the columns was added
up and the column with the highest score became the final diagnosis. If 2 or more columns scored
equally, the algorithm looks into specific features, else the diagnosis will remain inconclusive. The
genomic aspect of the tool included data extraction from .csv document files, wherein the patient
genome is compared with that of the reference genome. From this data, the tool prioritises genes that
are linked to the clinical disease computed by the matrix, followed by variant prioritization of exonic
(protein-coding) regions, filtering out synonymous variants and variants with a MAF>1%. The
remaining variants are compared with a pre-developed variant database containing all reported
variants for EB, thus enabling the identification of either reported or novel variants that would explain
the phenotype.

The developed online tool uses HTML, CSS with Jinja templating engine and React in the frontend and
Flask for the backend. Vamsi’s team created a PostgreSQL database for storing the input data and
used and Gunicorn "Green Unicorn", a Python Web Server Gateway Interface HTTP server for handling
requests. The major tasks achieved in the frontend were designing the layout and appearance of the
website, integrating it to the backend to handle requests and responses, defining routes for easier
navigation and handling pagination. For the layout and design they took inspiration from Ant Design
which is a UI toolkit for React, and used Fetch API to make API calls to our backend and get the required
data. The major tasks achieved in the backend were defining REST APIs, implementing resources and
models for the APIs and connecting it with the EBCDM logic written in Python. The automated online
version of the tool is developed at CSIR-IGIB and is currently deployed on GitHub.