Consulting services in the following areas:
- Data science
- CRISP-DM methodology
Data Science
Reducing human effort that is required to finalize the diagnosis
Overview
A client is a bioinformatician that works on Mendelian disease diagnostics using genomic analysis software.
Services
Challenge
The client has created a software that analyses genomic information using advanced filtering technology in order to help to diagnose Mendelian disease. The software reduces the list of possible pathogenic mutations; however, a further human effort is required in order to finalize the diagnosis.
Solution
Our data science team joined the effort in order to prove that machine learning based algorithms can output diagnosis without any human analysis. The team has performed data preparation and cleaning using 1k-genome open data set and labeled data set provided by the client, performed feature engineering with tight cooperation with a domain expert and implemented classifier based on Random Forest Classifier algorithm. Finally, the team provided the report of the task solution.
Results
In 2 weeks Lohika team:
- Prepared training and validation data sets
- Evaluated several ML algorithms for classification
- Determined the algorithm that works best for this case
- Trained a model using prepared data set
- Provided confusion matrix for the trained algorithm
As a result, the client had a proof that an ML-based algorithm can output the correct diagnosis.
