Rapid and robust sex determination from ancient enamel proteomes using protSexInferer

The authors developed protSexInferer, an automated, open-source bioinformatic pipeline that improves the accuracy and standardization of sex determination in ancient remains by utilizing a ratio-based approach (RAMELY) to analyze amelogenin peptides, thereby overcoming the limitations of manual verification and false-positive signals in degraded paleo-proteomic samples.

The Gist

Imagine you are an archaeologist digging up an ancient tooth. You want to know: Was this person a man or a woman?

In the past, you had two main ways to guess:

1. Look at the bones: If the skull or pelvis is missing or too broken, you're stuck.
2. Read the DNA: If the DNA is too old or degraded (like a shredded book), you can't read the story.

But teeth are tough! They are like nature's steel vaults, protecting proteins inside them for thousands of years. Scientists have been using these proteins to guess the sex, but the old method was like trying to solve a puzzle by looking at just one single piece. If that one piece looked a little bit like a "male" piece by mistake (a false alarm), the whole puzzle was ruined, and researchers had to spend hours manually checking every single clue to be sure.

Enter: protSexInferer (The "Smart Sex Detective")

This new paper introduces a brand-new tool called protSexInferer. Think of it as a super-smart, automated robot detective that doesn't just look for one clue, but looks at the whole crowd of clues to make a decision.

Here is how it works, using simple analogies:

1. The "Two Teams" Analogy

Inside every tooth, there are two types of protein messages (peptides) related to sex:

• Team X (AMELX): Found in everyone (men and women).
• Team Y (AMELY): Found only in men.

The Old Way: The old method was like a bouncer at a club. If they saw even one person wearing a "Team Y" shirt, they shouted, "It's a man!" But sometimes, a Team X person might wear a shirt that looks like a Team Y shirt by accident (a false positive). The bouncer would get confused, and the researcher would have to manually check every single shirt to see if it was real or a fake.

The New Way (protSexInferer): This tool counts everyone in the room.

• It counts how many "Team Y" shirts it sees.
• It counts how many "Team X" shirts it sees.
• Then, it calculates a Ratio (The "RAMELY" score).

If you have 100 Team X shirts and 0 Team Y shirts, the ratio is 0. That's a Woman.
If you have 100 Team X shirts and 50 Team Y shirts, the ratio is high. That's a Man.

2. Why is this better?

Imagine you are at a noisy party.

• The Old Method: You hear one voice that might be a man's voice. You aren't sure. You have to stop the music, walk over, and ask, "Are you a man?" (This is the tedious manual checking).
• The New Method: You look at the whole crowd. Even if one person is wearing a fake "Man" costume (a false alarm), the tool looks at the balance. If 99% of the voices are clearly "Women" and only 1% are "Men," the tool knows, "Okay, that one weird voice is just noise. This is still a woman's party."

Because it uses a ratio (a percentage) instead of just "Yes/No," it ignores the noise and the mistakes. It's like using a weather forecast that looks at the whole map, rather than just one cloud.

3. The "Magic Box" (The Pipeline)

The authors built this tool as a "pipeline." Imagine a factory assembly line:

1. Input: You dump in your raw data (the messy list of protein clues from the tooth).
2. Processing: The robot sorts the clues, throws away the duplicates, and ignores the fake ones.
3. Calculation: It does the math to find the ratio.
4. Output: It spits out a clear report: "This tooth is Male" or "This tooth is Female," along with a confidence score (like "95% sure").

Why does this matter?

• It's Fast: No more hours of staring at spreadsheets.
• It's Robust: It works even on tiny, broken, or very old teeth where the signal is weak.
• It's Honest: It admits when it doesn't know (if the data is too messy), rather than guessing wrong.
• It's Open: Anyone can use it for free.

The Bottom Line

This paper gives archaeologists a superpower. They can now take a tiny, ancient tooth fragment, run it through this "Smart Detective" software, and get a reliable answer about the person's sex in minutes, without needing perfect DNA or a perfect skeleton. It turns a messy, confusing puzzle into a clear, automated solution.

Technical Summary

1. Problem Statement

Accurate biological sex determination of ancient remains is critical for archaeological, anthropological, and forensic research. However, current methods face significant limitations:

• Morphological Analysis: Often unreliable for subadults, fragmented remains, or specimens lacking sexually dimorphic skeletal features.
• Ancient DNA (aDNA): Requires well-preserved endogenous DNA, which is frequently absent or degraded in ancient samples.
• Current Paleo-proteomic Workflows: Existing methods rely on detecting specific sexually dimorphic amelogenin peptides (AMELX on the X chromosome and AMELY on the Y chromosome).
  • Manual Verification: Current workflows depend on the manual inspection of a few specific peptide markers (e.g., AMELY-59M vs. AMELX-60) to confirm sex.
  • False Positives: Mass spectrometry database searches often yield false-positive AMELY signals in female samples due to homologous regions or database search artifacts.
  • Lack of Standardization: The reliance on manual verification makes the process non-standardized, time-consuming, and difficult to scale, limiting reproducibility.

2. Methodology: protSexInferer

The authors developed protSexInferer, a lightweight, open-source bioinformatic pipeline (written in Nextflow) designed to automate sex estimation from paleo-proteomic data.

Core Workflow:

1. Input: Accepts output files from major protein search engines (PEAKS, pFind, MaxQuant, and DIA-NN).
2. Reference Databases: Utilizes pre-constructed, optimized protein reference databases (e.g., "Hominidae enamel protein database") containing amelogenin isoforms (AMELX and AMELY) and other enamel proteins. Users can also input custom databases.
3. Peptide Classification:
   • Parses search results to extract amelogenin peptides.
   • Filters peptides to identify those unique to AMELY or AMELX.
   • Crucial Step: Excludes peptides that match both AMELX and AMELY sequences (homologous regions) to ensure only truly specific peptides are quantified.
4. RAMELY Calculation:
   • Instead of relying on the mere presence/absence of AMELY signals, the pipeline calculates the RAMELY ratio:
     $$RAMELY = \frac{n_{AMELY}}{n_{AMELY} + n_{AMELX}}$$
     Where $n$ represents the count of unique, filtered peptides for each isoform.
   • Computes 95% confidence intervals (CI) based on a Bernoulli distribution assumption.
5. Sex Assignment:
   • Compares the sample's RAMELY value against established thresholds derived from known-sex reference samples.
   • Female: $RAMELY \leq$ Upper limit of 95% CI of max female reference.
   • Male: $RAMELY \geq$ Lower limit of 95% CI of min male reference.
   • Unknown: Values falling between thresholds or overlapping boundaries.
6. Output: Generates comprehensive tables of peptide sequences, intensity data, RAMELY values with CIs, and a final sex determination report.

3. Key Contributions

• Ratio-Based Approach: Shifts the paradigm from binary signal detection (presence/absence) to a quantitative ratio (RAMELY), which effectively mitigates the impact of false-positive AMELY signals.
• Automation & Standardization: Provides an end-to-end, automated pipeline that eliminates the need for time-consuming manual verification of peptide markers.
• Multi-Engine Compatibility: Supports outputs from four major search engines (PEAKS, pFind, MaxQuant, DIA-NN), allowing for cross-platform validation.
• Robustness to Low Data: Demonstrates reliability even in samples with low peptide yields or where specific diagnostic peptides (like 59M) are missing, by utilizing the full spectrum of available AMELX/AMELY peptides.
• Open Source: The pipeline and reference databases are publicly available on GitHub, promoting reproducibility.

4. Results

• Validation on Reference Data: The pipeline was tested on 76 reference samples of known sex. The RAMELY values formed two distinct clusters corresponding to biological sex, regardless of the search engine used.
  • Thresholds: PEAKS was selected as the default engine due to the largest separation between male and female clusters (Male threshold: $>0.088$; Female threshold: $<0.055$).
• Independent Validation: Applied to 69 independent samples (including ancient, modern, adult, and subadult specimens). The pipeline's assignments matched prior determinations in almost all cases.
  • Reclassification: Successfully reclassified individual JY63 (previously morphologically male) as female, aligning with previous proteomic findings, demonstrating the method's ability to correct morphological errors.
• Handling False Positives:
  • Traditional intensity-based methods (comparing AMELY-59M vs. AMELX-60) frequently misclassified females as males due to false-positive AMELY signals.
  • The RAMELY ratio approach correctly identified these females by showing that the proportion of AMELY peptides remained within the female threshold range, despite the presence of some false-positive signals.
• Low-Abundance Samples: The method remained effective for samples with very low AMELY signal intensity or missing typical diagnostic peptides, proving robustness for highly degraded ancient DNA-free specimens.
• New Application: Successfully determined the sex of 8 scattered teeth from the Xiawanggang site (~5,000 BP), classifying them all as male with high confidence.

5. Significance

• Enhanced Reliability: By utilizing a ratio-based metric rather than binary detection, protSexInferer significantly reduces false-positive rates, a major bottleneck in paleo-proteomic sexing.
• Scalability: The automated nature of the pipeline allows for the rapid processing of large-scale archaeological datasets, which was previously unfeasible with manual workflows.
• Broad Applicability: The method is effective across diverse sample types (adults, subadults, deciduous/permanent teeth) and preservation states (up to ~2 million years old), making it a versatile tool for paleoanthropology and forensics.
• Standardization: Establishes a standardized, reproducible framework for the field, facilitating cross-study comparisons and meta-analyses of ancient population structures.

In conclusion, protSexInferer represents a major advancement in paleo-proteomics, transforming sex determination from a subjective, manual process into a robust, automated, and statistically rigorous workflow.