Non-Invasive Paternity & Kinship Analysis in Wild Mountain Gorillas

---

Research Question

Does reproductive success map strictly onto male dominance rank in a structured primate society? Specifically, do dominant silverbacks monopolise all reproduction, or do subordinate males contribute meaningfully to paternity in wild mountain gorilla groups?

---

Key Findings Summary

1. On paternity, 60% of offspring were sired by the dominant silverback, 10% by subordinate males, and 30% were unresolved; likely extra-group males. So dominants did not monopolise reproduction. Subordinate paternity was real and detected across multiple groups, which directly challenges strict reproductive skew models.

2. On reproductive skew, the Nonacs B-index came out above the equal-sharing expectation, confirming that rank does confer an advantage. But the skew ratio was nowhere near the monopolisation extreme. The permutation test told you whether that skew was statistically distinguishable from random; and that result is data-dependent, so yours will vary slightly each run.

3. On kinship, within-group relatedness was significantly higher than between-group relatedness (Mann-Whitney p < 0.05), which is exactly what you'd expect from philopatric females and resident males. The between-group signal was low but non-zero, consistent with occasional male dispersal acting as natural gene flow across the Virunga sub-massifs.

4. On genetic diversity, mean expected heterozygosity across 15 loci was around 0.65–0.70 with mean PIC above 0.5, meaning the locus panel was highly informative. No group showed catastrophic diversity collapse, but the conservation notebook flags groups falling below He = 0.6 as warranting monitoring.

5. The headline conclusion: dominant silverbacks have a real reproductive advantage, but the system is not winner-takes-all. Subordinate paternity is a persistent feature, not noise; which has direct implications for group management, because removing subordinates based on rank alone would cost you genetic diversity you didn't know they were contributing.

Background

Mountain gorillas live in structured social groups with clear male dominance hierarchies. The conventional expectation is that dominant silverbacks monopolize reproduction (high reproductive skew). This pipeline tests that assumption rigorously using microsatellite-based parentage analysis and pairwise kinship estimation.

Key biological parameters modelled:

• 6 social groups across 3 Virunga sub-massifs (Karisimbi, Bisoke, Sabyinyo/Muhabura)
• 15 microsatellite loci (panel modelled on Bradley et al. 2004; Nsubuga et al. 2010)
• Non-invasive DNA error structure: allelic dropout ~8%, false alleles ~2%, locus failure ~7%
• Ground-truth paternity split: dominant ≈60%, subordinate ≈10%, extra-group ≈30%

---

Tools: Python · R (via Jupyter) · CERVUS-style LOD scoring · Microsatellite Genotyping

Alternative Dataset

Gorilla microsatellite data:

• Dryad: https://datadryad.org , search "Gorilla beringei microsatellite"
• Nsubuga et al. 2010 (Bwindi gorillas): Mol Ecol Resour, 10(2):397–399
• Bradley et al. 2004 (Virunga gorillas): Am J Primatol, 62(1):1–14
• NCBI PopSet: https://www.ncbi.nlm.nih.gov/popset , "search gorilla STR"

Data Schema

File	Key Columns
individuals.csv	individual_id, name, sex, age_class, group_id, dominance_rank, mother_id, true_father_id
genotypes.csv	individual_id, {LOCUS}_a1, {LOCUS}_a2 (15 loci × 2 alleles each)
groups.csv	id, name, location

---

Methods Summary

Parentage Analysis (CERVUS-style LOD scoring)

Following Marshall et al. (1998) Mol Ecol 7:639–655:

$$\text{LOD} = \sum_{l=1}^{L} \log_{10} \frac{P(G_o^l \mid G_m^l, G_c^l)}{P(G_o^l \mid G_m^l)}$$

Where $G_o, G_m, G_c$ are offspring, mother, and candidate father genotypes. Candidate with highest positive LOD and delta-LOD > 0 is assigned paternity.

Kinship / Relatedness

• Queller & Goodnight (1989) rxy estimator
• Lynch & Ritland (1999) r estimator
• Proportion of alleles shared (IBS) as a non-parametric check

Reproductive Skew

Nonacs B-index (Nonacs 2000, Am Nat):

$$B = \sum_{i=1}^{k} \left(p_i - \frac{1}{k}\right)^2$$

Compared to null distribution via permutation (1000 replicates).

---

References

1. Marshall TC et al. (1998) Statistical confidence for likelihood-based paternity inference. Mol Ecol 7:639–655.
2. Bradley BJ et al. (2004) Dispersal patterns, phenotype, and genotype in free-ranging gorillas. Am J Primatol 62:1–14.
3. Nsubuga AM et al. (2010) Gorilla microsatellite loci. Mol Ecol Resour 10:397–399.
4. Vigilant L et al. (2001) Paternity and relatedness in wild chimpanzee communities. PNAS 98:12890–12895.
5. Nonacs P (2000) Measuring and using skew in the study of social behaviour. Am Nat 156:577–589.
6. Queller DC, Goodnight KF (1989) Estimating relatedness using genetic markers. Evolution 43:258–275.
7. Lynch M, Ritland K (1999) Estimation of pairwise relatedness. Genetics 152:1753–1766.

---

Project Structure

gorilla-paternity-analysis/
├── README.md
├── .gitignore
├── requirements.txt
├── environment.yml
├── data/
│   ├── raw/                    
│   └── processed/              
├── notebooks/
│   ├── 01_data_import_and_qc.ipynb
│   ├── 02_population_genetics.ipynb
│   ├── 03_parentage_analysis.ipynb
│   ├── 04_kinship_analysis.ipynb
│   └── 05_results_and_conservation.ipynb
├── scripts/
│   ├── parentage_functions.py
│   ├── kinship_functions.py
│   └── utils.py
└── outputs/
    ├── figures/
    └── tables/

Author

Data Technician with research experience at the Dian Fossey Gorilla Fund, building end-to-end computational pipelines across four domains: spatial movement ecology (GeoPandas, KDE, permutation testing), population genetics (CERVUS microsatellite LOD scoring, Queller-Goodnight kinship estimation), machine learning survival analysis (Random Forest, temporal cross-validation), and conservation epidemiology (logistic regression, SciPy hypothesis testing, temporal linkage). Technical stack: Python · R · SQL · scikit-learn · SciPy · GeoPandas · Git. All work is grounded in longitudinal biological datasets with direct conservation policy implications across the Virunga Massif — Rwanda, Uganda, and DRC.