baolef/rna-active-learning

cmu capstone project: active learning with rna

rna-active-learning

Capstone project for 02763 MSAS Capstone 1 at CMU

This project solves multiple rna related problem using xgboost with active learning.

Environment

To set up the environment, you can run

conda env create -f environment.yml

To activate the conda environment, you can run

conda activate 02763

Data

There are two types of data. One describes the class of each rna sequence. The other describes a function based on rna features.

Class labels

To extract classes from rna sequences, you can run

python data/classifier.py

Regression labels

To generate a regression object from rna sequences, you can run

python data/prediction.py

Training

There are three models solving classification problem. The

Class prediction

To predict the class of the rna sequence, with the level (eg. 3) of class we want to predict, we can run

python models/classifier.py -i inputs/SILVA_138_3_8_sliding_0 -c 3

Note here active learning is not implemented due to computation complexity.

Class pairing

To predict whether a pair of rna sequences comes from the same class (eg. level 3), you can run

python models/pair.py -i inputs/SILVA_138_3_8_sliding_0 -c 3

Function predicting

To predict the function, which is also formed as a classification problem by setting a threshold, we can run

python models/prediction.py -i inputs/SILVA_138_3_8_sliding_0 -p 80