pkpinto / tfnet

Study of transcriptor factor (TF) binding site networks

TFnet

There are several notebooks in this folder. They were created to do preliminary studies on the transciptor factor (TF) binding site data of the Uniprobe database. Uniprobe contains data on 21 species (with different kinds of coverage). This data consists of the binding streght of a TF to all kmers of length 8. The data is available as a mySQL dump (complete dataset) or as zipped files containing partial dataset (including only specific species and/or TFs).

uniprobe_data.ipynb

The SQL dump is provided without any ER diagram or documentation. This notebook imports the database (after conversion into a SQLite database) for exploratory discovery and conversion to a more intelligible format. Some preprocessing is done directly in SQL (stuff like merging identical tables and such). Data which does not fit our needs (e.g. TF complexed with proteins) was also selected and discarded.

uniprobe_data.py

Python library that makes the database available to other programs. Use with

from uniprobe_data import *

Other notebooks

Additionally, other notebooks are dedicated to more detailed analysis.

bind_graphs.ipynb

Study the TF binding sites per TF. Create binding sites networks for each of the TFs. Do TF pair-wise comparisons of the sets of binding sites.

uniprot_seq.ipynb

For the TF in the Uniprobe database, mine Uniprot and PFAM for sequence and domain data and add it to the database. Make TF pair-wise comparisons.

clustering.ipynb

Cluster TFs according to pairwise comparison statistics: binding site sets similarity, sequence similarity and BLAST score and domain content.

Notes

What is shown here is the exploratory phase of the project. Looking into the data, understanding what is and what is not usable and making it available for further analysis.