README_OLD_VERSION.md

metadata

setup: bash setup.sh
title: MtDNALocation
emoji: 📊
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 5.25.0
app_file: app.py
pinned: false
license: mit
short_description: mtDNA Location Classification tool

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

Installation

Set up environments and start GUI:

git clone https://github.com/Open-Access-Bio-Data/mtDNA-Location-Classifier.git

If installed using mamba (recommended):

mamba env create -f env.yaml

If not, check current python version in terminal and make sure that it is python version 3.10, then run

pip install -r requirements.txt

To start the programme, run this in terminal:

python app.py

Then follow its instructions

Descriptions:

mtDNA-Location-Classifier uses Gradio to handle the front-end interactions.

The programme takes an accession number (an NCBI GenBank/nuccore identifier) as input and returns the likely origin of the sequence through classify_sample_location_cached(accession=accession_number). This function wraps around a pipeline that proceeds as follow:

Steps 1-3: Check and retrieve base materials: the Pubmed ID, isolate, DOI and text:

• Which are respectively:

Step 1: pubmed_ids and isolates

    `get_info_from accession(accession=accession_number)`
- Current input is a string of `accession_number` and output are two lists, one of PUBMED IDs and one of isolate(s).
- Which look through the metadata of the sequence with `accession_number` and extract `PUBMED ID` if available or `isolate` information.
- The presence of PUBMED ID is currently important for the retrieval of texts in the next steps, which are eventually used by method 4.1 (question-answering) and 4.2 (infer from haplogroup)
- Some sequences might not have `isolate` info but its availibity is optional. (as they might be used by method 4.1 and 4.2 as alternative)

Step 2: dois

    `get_doi_from_pubmed_id(pubmed_ids = pubmed_ids)`
- Input is a list of PUBMED IDs of the sequence with `accession_number` (retrieved from previous step) and output is a dictionary with keys = PUBMED IDs and values = according DOIs.
- The pubmed_ids are retrieved from the `get_info_from accession(accession=accession_number)` mentioned above.
- The DOIs will be passed down to dependent functions to extract texts of publications to pass on to method 4.1 and 4.2

Step 3: get text

    `get_paper_text(dois = dois)`
- Input is currently a list of dois retrieved from previous step and output is a dictionary with keys = sources (doi links or file type) (We might improve this to have other inputs in addition to just doi links - maybe files); values = texts obtained from sources.
- Output of this step is crucial to method 4.1 and 4.2

Step 4: Prediction of origin:

Method 4.0:

- The first method attempts to directly look in the metadata for information that was submitted along with the sequence. Thus, it does not require availability of PUBMED IDs/DOIs or isolates.
- However, this information is not always available in the submission. Thus, we use other methods (4.1,4.2) to retrieve publications through which we can extract the information of the source of mtDNA

Method 4.1:

- 

Method 4.2:

- 

More in the package

extraction of text from HTML

extraction of text from PDF