Pubnomics | Devpost

Small region of BRCA1. Hovering over an amino acid will tell you its position and type.
For users who aren't familiar with the variant nomenclature, there is the option to click and select the resulting protein.
Currently we aggregate data about a mutation from 3 public data sources: ClinVitae, ClinVar, and ARUP.

Inspiration

Currently solutions for learning more about genetic mutations is inaccessible and hard to digest.

What it does

Pubnomics aggregates trusted public data about the pathogenicity of genetic mutations and displays it in a simple visual-driven interface.

How I built it

My partner worked on the frontend using React.js while I built the backend using Go. The app is hosted on Google cloud platform's Compute Engine and leverages a self-hosted CockroachDB instance for storing genomics data. We ran a dataset from UMD's BRCA1 database to train the variant prediction algorithm.

Challenges I ran into

Visualizing Genes is hard because they are extremely lengthy concepts, with each base pair being very important. Building an interface for scientists meant that everyone was a power user, and therefore we had a very small room for latency. We reviewed every React component to try and squeeze performance using the shouldComponentUpdate optimization.

Fetching data from multiple domains was another tricky problem we solved. Most sites have a strict cross domain policy, so our initial idea of loading the site and parsing it through jQuery was impossible. Instead, we wrote a server hosted on a VM on the Google cloud platform to fetch it via an http client included in the Go stdlib. Since we wrote the server in Go, fetching multiple data sources in parallel was trivial, thus reducing the amount of latency the user has to experience to get important genetic information.

Ranking heuristics in human studies is hard. Putting values on human biases is a hard judgement call because you are susceptible to your own bias.