Skip to content

Chlamy Atlas is a AI-powered web application which predicts the localizations of proteins from the Green Algae Chlamydomonas reinhardtii.

Notifications You must be signed in to change notification settings

CSBiology/ChlamyAtlas

Repository files navigation

ChlamyAtlas

A web UI for optimised versions of the models published in Wang et al. 2023.

alt text

Supported formats

ChlamyAtlas expects input in either the FASTA format or as pure amino acid sequence. The FASTA format consists of two building blocks. The first is a description which explains the following sequence. This description starts with ">" and is written in a single line. The amino acid sequence follows in the next line and can span multiple lines. An example for this format is:

>sp|A0A178WF56|CSTM3_ARATH Protein CYSTEINE-RICH TRANSMEMBRANE MODULE 3 OS=Arabidopsis thaliana OX=3702 GN=CYSTM3 PE=1 SV=1
MAQYHQQHEMKQTMAETQYVTAPPPMGYPVMMKDSPQTVQPPHEGQSKGSGGFLRGCLAA
MCCCCVLDCVF
>sp|A1YKT1|TCP18_ARATH Transcription factor TCP18 OS=Arabidopsis thaliana OX=3702 GN=TCP18 PE=1 SV=1
MNNNIFSTTTTINDDYMLFPYNDHYSSQPLLPFSPSSSINDILIHSTSNTSNNHLDHHHQ
FQQPSPFSHFEFAPDCALLTSFHPENNGHDDNQTIPNDNHHPSLHFPLNNTIVEQPTEPS
ETINLIEDSQRISTSQDPKMKKAKKPSRTDRHSKIKTAKGTRDRRMRLSLDVAKELFGLQ
DMLGFDKASKTVEWLLTQAKPEIIKIATTLSHHGCFSSGDESHIRPVLGSMDTSSDLCEL
ASMWTVDDRGSNTNTTETRGNKVDGRSMRGKRKRPEPRTPILKKLSKEERAKARERAKGR
TMEKMMMKMKGRSQLVKVVEEDAHDHGEIIKNNNRSQVNRSSFEMTHCEDKIEELCKNDR
FAVCNEFIMNKKDHISNESYDLVNYKPNSSFPVINHHRSQGAANSIEQHQFTDLHYSFGA
KPRDLMHNYQNMY

ChlamyAtlas was developed with the assumption that the description follows the standard used by the Universal Protein Resource (Uniprot) and only returns the Uniprot ID as description in the output table. This can be circumvented by removing the "|" in the description. In this case the complete description gets returned.

The only other supported format are pure amino acid sequences. An example for this format is:

MAQYHQQHEMKQTMAETQYVTAPPPMGYPVMMKDSPQTVQPPHEGQSKGSGGFLRGCLAA
MCCCCVLDCVF

This format can only be used for a single amino acid sequence. Multiple amino acid sequences must be in the following format:

>!MAQYHQQHEMKQTMAETQYVTAPPPMGYPVMMKDSPQTVQPPHEGQSKGSGGFLRGCLAA
MCCCCVLDCVF
>!MNNNIFSTTTTINDDYMLFPYNDHYSSQPLLPFSPSSSINDILIHSTSNTSNNHLDHHHQ
FQQPSPFSHFEFAPDCALLTSFHPENNGHDDNQTIPNDNHHPSLHFPLNNTIVEQPTEPS
ETINLIEDSQRISTSQDPKMKKAKKPSRTDRHSKIKTAKGTRDRRMRLSLDVAKELFGLQ
DMLGFDKASKTVEWLLTQAKPEIIKIATTLSHHGCFSSGDESHIRPVLGSMDTSSDLCEL
ASMWTVDDRGSNTNTTETRGNKVDGRSMRGKRKRPEPRTPILKKLSKEERAKARERAKGR
TMEKMMMKMKGRSQLVKVVEEDAHDHGEIIKNNNRSQVNRSSFEMTHCEDKIEELCKNDR
FAVCNEFIMNKKDHISNESYDLVNYKPNSSFPVINHHRSQGAANSIEQHQFTDLHYSFGA
KPRDLMHNYQNMY

Result

alt text

Explanations of Chloropred ,Qchloro, Mitopred,Qmito,Secrpred,Qsecr, and FinalPred.

Chloropred

Prediction score indicating the likelihood of the protein being localized to the Chloroplast. A higher scores suggest a stronger prediction that the protein is localized in the Chloroplast.

Qchloro

q-value associated with the Chloroplast prediction score. Provides a measure of statistical significance for the Chloroplast prediction. Lower q-values indicate higher statistical significance.

Mitopred

Prediction score for the localization of the protein to the Mitochondria. A higher scores suggest a stronger prediction of Mitochondrial localization.

Qmito

q-value associated with the Mitochondria prediction score. Indicates the statistical significance of the Mitochondria localization prediction. Lower q-values suggest a more reliable prediction.

Secrpred

Prediction score for identifying the protein as a Secretory Protein.A higher scores indicate a stronger likelihood that the protein functions as a Secretory Protein.

Qsecr

q-value for the Secretory Protein prediction. Provides a measure of the statistical significance of the Secretory Protein prediction. Lower q-values are indicative of more statistically significant predictions.

FinalPred

Represents the model's final prediction of the protein's localization based on the highest score and its corresponding q-value. The final localization is determined by comparing the q-values and prediction scores against preset cutoffs. If all q-values exceed the cutoff, the protein is classified as "Cytoplasmic."

Cutoff

The threshold q-value below which a prediction is considered statistically significant. Set to 0.05 by default, meaning that predictions with q-values below this threshold are classified as significant. This parameter helps in distinguishing between statistically significant and non-significant predictions, reducing the chance of false-positive localizations.

Docker

Environment Variables

  • NET_EMAIL_EMAIL: Email address to send emails from

    Default: Set via user secrets

  • NET_EMAIL_ACCOUNTNAME: Email account name to send emails from

    Default: Set via user secrets

  • NET_EMAIL_PASSWORD: Email account password to send emails from

    Default: Set via user secrets

  • NET_EMAIL_SERVER: Email server to send emails from

    Default: Set via user secrets

  • NET_EMAIL_PORT: Email server port to send emails from

    Default: Set via user secrets

  • PYTHON_SERVICE_TIMEOUT: Time in minutes before conenction timeout between ui and api service

    Default: 30 minutes

  • PYTHON_SERVICE_URL: Sets the url for the api predictor backend.

    Default: http://localhost:8000

    Remarks: In docker compose this could be http://host.docker.internal:8000

    Remarks: On Linux might require:

    extra_hosts:
        - "host.docker.internal:host-gateway"
    
  • PYTHON_SERVICE_STORAGE_TIMESPAN: How long the user data should be stored

    Default: 1 Hour

Docker Compose

version: '3.7'

name: chlamyatlas

services:
  api:
    image: csbdocker/chlamyatlas-api:latest
    ports:
      - 8000:80
    environment:
      GUNICORN_CMD_ARGS: "-k uvicorn.workers.UvicornWorker --preload"
      MAX_WORKERS: "4"
      TIMEOUT: "0"
  ui:
    image: csbdocker/chlamyatlas-ui:latest
    environment:
      PYTHON_SERVICE_URL: "http://host.docker.internal:8000"
      PYTHON_SERVICE_STORAGE_TIMESPAN: "7"
    ports:
      - 5000:5000
    # Use this to make host.docker.internal accessible on linux docker
    extra_hosts:
     - "host.docker.internal:host-gateway"

Local Development

Install pre-requisites

You'll need to install the following pre-requisites in order to build SAFE applications

Install

  • run setup.cmd

.. or ..

  1. dotnet tool restore
  2. py -m venv .venv
  3. .\.venv\Scripts\python.exe -m pip install -r .\src\FastAPI\requirements.txt

Run

  • .\build.cmd run starts SAFE stack

plus in another terminal run:

  1. activate local python environment: .\.venv\Scripts\Activate.ps1
  2. navigate to fastapi folder: cd .\src\FastAPI\
  3. start fastapi backend: ./run.cmd

Activate Email notification (optional)

Set user-secrets in the following schema:

{
  "email": {
    "NET_EMAIL_EMAIL": "[email protected]",
    "NET_EMAIL_ACCOUNTNAME": "PlaceholderAccountName",
    "NET_EMAIL_PASSWORD": "HelloWorld1234",
    "NET_EMAIL_SERVER": "smtp.placeholdermail.de",
    "NET_EMAIL_PORT": 587
  }
}

Publish

Test Publish

  1. .\build.cmd dockerbundle [--uionly], creates :new docker image(s). Skip fastapi image with --uionly
  2. .\build.cmd dockertest, uses local docker-compose file to start :new images.

To docker-hub

  1. Login to CSB-Docker
  2. Ensure correct Versions, both for python and dotnet service.
    • .\build.cmd versions
    • Remarks: Versions are defined in project files. Paths can be found in build project ProjectInfo.fs. Accessed via regex parsing.
  3. Run Test Publish steps. The following step requires built :new images.
  4. .\build.cmd dockerpublish

Request Workflow

sequenceDiagram
    participant py as Python ML
    participant net as F#35; Server
    participant c as Client
    actor u as User
    u -->> c: Gives data
    c -->>+net: sends user data
    par start analysis
    loop
    net-)+py: send sequence
    py->py: predict target
    py-)net: return predicted target
    end
    and return request information
    net -) c: returns `request-ID`
    end
    critical ⚠️
    u -->> c: copies and stores `request-ID`
    end
    opt email
    u -->> c: give email address
    c -->> net: give id + email to store
    end
    opt check status
    u -->> c: use `request-ID` to check status
    end
    py-)net: send last package
    deactivate py
    net-->>net: run q-value calculation
    net-->>net: store results
    deactivate net
    opt gave email
    net-)u: send email
    end
    u -->> c: request data
    c-->>net: get data
    net-->>c: return data
    c-->>u: download data
Loading