Forbidden Words: Nullomer Constraints on ABC Transporter Regulatory Evolution

Harrizi S., Nait Irahal I., Kabine M.
Laboratoire Santé, Environnement et Biotechnologie (LSEB), Faculté des Sciences Ain Chock, Université Hassan II de Casablanca, Morocco.

---

Overview

This repository contains the complete analysis pipeline for:

Forbidden Words: How Nullomer Constraints Channel Regulatory Evolution in ABC Transporters

We identify 463,220 absent 11-mer sequences (nullomers) in the S. cerevisiae S288C genome and measure the evolutionary constraints they impose on 26 ABC transporter genes and their promoters. Key findings:

Result	Value
Nullomers identified (k=11)	463,220 (11.04% of theoretical)
Nullomer mean GC content	65.7% vs 38.3% genome-wide
Total NEMs across 26 genes	174,799
Promoter NEM density	1,331.9 NEMs/kb (22.6% enrichment over gene bodies, p=0.003)
PDRE–NEM correlation	Spearman ρ=0.685, p=1.1×10⁻⁴
Drug efflux vs other NEM density	+28% (p=0.018, Cohen's d=1.08)
Nullomer ΔG vs random	−13.96 vs −12.13 kcal/mol (ΔΔG=1.83, 19.4-fold disadvantage)
ML model (Random Forest)	Test R²=0.760, CV R²=0.717±0.045
GP fitness landscape	R²=0.896
Meta-analysis p-value	2.28×10⁻⁸

---

Repository structure

.
├── scripts/
│   ├── 01_nullomer_identification.py   # k=11 nullomers from S. cerevisiae genome
│   ├── 02_nem_analysis.py              # NEM identification across 26 ABC transporters
│   ├── 03_stress_element_analysis.py   # PDRE/STRE/HSE/AP-1 scanning and correlation
│   ├── 04_thermodynamic_analysis.py    # ΔG and Tm analysis (SantaLucia 1998)
│   ├── 05_ml_and_network_analysis.py   # Random Forest, GP landscape, STRING network
│   ├── 06_statistical_synthesis.py     # Hypothesis tests, dN/dS, positional analysis
│   └── 07_wright_fisher_simulation.py  # Wright-Fisher evolutionary simulations
├── docs/
│   ├── DATA_DICTIONARY.md
│   └── REPRODUCTION_GUIDE.md
├── run_all.sh                          # Runs all scripts in order
├── requirements.txt
└── README.md

Results are written to results/ and genome data is cached in data/.

---

Quickstart

git clone https://huggingface.co/datasets/HarriziSaad/Nullomer
cd Nullomer

pip install -r requirements.txt

# Run full pipeline (~60–90 min depending on hardware)
bash run_all.sh

# Or run individual steps
python scripts/01_nullomer_identification.py   # ~50 s
python scripts/02_nem_analysis.py              # ~30–60 min (NEM scanning)
python scripts/03_stress_element_analysis.py   # ~5 min
python scripts/04_thermodynamic_analysis.py    # ~2 min
python scripts/05_ml_and_network_analysis.py   # ~10 min
python scripts/06_statistical_synthesis.py     # ~2 min
python scripts/07_wright_fisher_simulation.py  # ~3 min

Scripts 03–07 depend on outputs of earlier scripts. Run in order, or run run_all.sh.

---

Data sources

All genomic data is downloaded automatically at runtime:

Resource	URL
S. cerevisiae genome (R64-1-1)	Ensembl release 110
Gene annotations (GFF3)	Ensembl release 110
Protein interactions	STRING v11.5, NCBI TaxID 4932, score ≥ 400

---

Do I need to upload results files?

No. All result files are generated by running the scripts. The results/ directory does not need to be uploaded to the repository. Each script documents exactly what it outputs.

---

Output files

File	Generated by
results/nullomers_k11.txt	01
results/nem_comprehensive_summary.csv	02
results/nem_enrichment_analysis.csv	02
results/stress_permutation_test.json	02
results/stress_element_nem_correlation.csv	03
results/motif_disruption_by_nems.csv	03
results/nullomer_thermodynamics.csv	04
results/thermodynamic_summary.json	04
results/ml_feature_importance.csv	05
results/ml_model_performance.json	05
results/network_topology.csv	05
results/fragility_scores.csv	05
results/statistical_synthesis.json	06
results/network_communities.csv	05
results/wf_trajectories.csv	07
results/wf_final_frequencies.csv	07
results/wf_simulation_summary.json	07

---

ABC transporters analyzed

The 26 genes span drug efflux pumps (PDR5, SNQ2, YOR1, PDR10, PDR11, PDR12, PDR15, PDR18, YCF1), transcriptional regulators (PDR1, PDR3, PDR16, PDR17), mitochondrial transporters (ATM1, MDL1, MDL2), translation-related proteins (YEF3, GCN20, ARB1, RLI1), and others (VMR1, YBT1, BPT1, HMT1, NMD5, STE6).

---

Citation

Harrizi S., Nait Irahal I., Kabine M. (2025). Forbidden Words: How Nullomer
Constraints Channel Regulatory Evolution in ABC Transporters.

---

License

Code: MIT
Data downloaded from Ensembl and STRING is subject to their respective terms of use.