Pioneer Factor Cohesin Chromatin Binding Predictor with PyTorch
Pioneer transcription factors direct cohesin loop extrusion across genomic chromatin boundaries. You will build a deep learning neural network in Python using PyTorch to predict cohesin binding sites from pioneer factor ChIP-seq motifs, achieving an AUROC of 0.91.
What you are building
Train a PyTorch convolutional model to predict pioneer-factor-guided cohesin chromatin binding from genomic sequence motifs.
How work is evaluated
Evaluates cohesin binding prediction AUROC and model loss on unseen genomic test chromosomes.
Datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
probability_range_check
Ensure output is a valid probability float between 0 and 1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
auroc
Area under the receiver operating characteristic curve on chromosome 21 test set • target: 0.91 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Learning goals
Build 1D CNN residual neural architectures in PyTorch for genomic sequence processing
Process ChIP-seq peak signals and pioneer factor binding motifs into tensor representations
Predict chromatin loop anchor sites and cohesin deposition probabilities
Resources
ChIP-seq genomic dataset covering 50,000 pioneer factor and cohesin binding region windows across human cell lines.
Challenge setup
Evaluates cohesin binding prediction AUROC and model loss on unseen genomic test chromosomes.
Challenge input
JSON object with dna_sequence string (1000bp)
PyTorch
Mandatory deep learning neural network framework.
Evaluated output
JSON object with binding_probability float
- Ensure output is a valid probability float between 0 and 1
- Area under the receiver operating characteristic curve on chromosome 21 test set • target: 0.91 • range: 0-1
- Benchmark: GenoDeep-Bench
- Auroc target: 0.91
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Shows the first task, up to three tools, and up to three evidence items.
View technical recipe
Configured tools
- Llama 3.2 · Required
- Mindgard · Optional
- LocalAI · Optional
Evaluation contract
- probability_range_check · Weight 1
- auroc · Weight 1
Metrics
- auroc · Target 0.91
Benchmarks
- GenoDeep-Bench
- biotech
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with dna_sequence string (1000bp)
Output: JSON object with binding_probability float
Execution harness: Python
Evaluation tools
- PyTorch · Deep Learning Framework
Recipe freeze marker
No recipe freeze time is recorded.