Feeding Route and Oral Microbiome in Children With Complex Chronic Conditions: 16S rRNA Sequencing and Processed Microbiome Data

# Feeding Route and Oral Microbiome in Children With Complex Chronic Conditions

## Repository scope

This Zenodo record accompanies the study **“Feeding Route and Oral Microbiome in Children With Complex Chronic Conditions”** and provides a curated, non-exhaustive set of raw and processed data products supporting traceability and reproducibility of the reported 16S rRNA gene microbiome analyses.

**Zenodo DOI:** https://doi.org/10.5281/zenodo.19020777

The repository prioritizes the original sequencing data, audited QIIME 2 artifacts, a minimal pseudonymized study metadata table, and processed functional data directly linked to the analyses described in the manuscript. Historical, exploratory, redundant, or retrospectively reconstructed outputs are intentionally excluded.

## Study overview

- **Design:** Observational study.
- **Setting:** La Paz University Hospital, Madrid, Spain.
- **Study period:** July–November 2023.
- **Participants:** 63 children and adolescents aged 5–17 years with complex chronic conditions.
- **Final feeding groups:** Oral feeding (n = 29), exclusive enteral feeding (n = 20), and mixed feeding (n = 14).
- **Oral microbiome samples:** 221.
- **Sequencing target:** V3–V4 region of the 16S rRNA gene.
- **Sequencing:** Illumina paired-end sequencing (250-bp reads).
- **Primary bioinformatics:** QIIME 2 (2022.2 series) with DADA2 for ASV inference and SILVA 138.1 for taxonomic assignment.
- **Predicted functional profiling:** PICRUSt2 with KEGG Ortholog-based functional reconstruction and participant-level KEGG Level 3 pathway summaries.

Detailed participant eligibility, oral sampling procedures, laboratory methods, and statistical analyses are described in the associated manuscript.

## Files in this repository

### Raw sequencing data

Raw FASTQ files are already present in the Zenodo record. They should be retained in subsequent versions and are not duplicated in the processed-data additions listed below.

### QIIME 2 artifacts

#### `denoisingTable.qza`

QIIME 2 `FeatureTable[Frequency]` artifact generated by the DADA2 `denoise_single` step.

Audited properties:
- 221 samples
- 5,600 features
- 11,992,865 total feature counts
- QIIME 2 framework 2022.2.1
- DADA2 plugin 2022.2.0
- embedded QIIME 2 provenance retained

This artifact preserves the original hashed feature identifiers generated at the denoising stage.

#### `featureTable.qza`

QIIME 2 `FeatureTable[Frequency]` artifact containing the processed ASV feature table used downstream.

Audited properties:
- 221 samples
- 5,600 features
- 11,992,865 total feature counts
- ASV identifiers `ASV0`–`ASV5599`

After sample alignment, its feature-count information is quantitatively concordant with `denoisingTable.qza`, while the downstream feature identifiers differ.

#### `featureSeqs.qza`

QIIME 2 `FeatureData[Sequence]` artifact containing 5,600 representative ASV sequences.

The feature identifiers are `ASV0`–`ASV5599` and match the identifiers in `featureTable.qza`.

#### `seq_taxonomy.qza`

QIIME 2 `FeatureData[Taxonomy]` artifact containing 5,600 taxonomic annotations.

The feature identifiers match `featureTable.qza` and `featureSeqs.qza` one-to-one.

### Public metadata

#### `study_metadata_public.csv`

Minimal pseudonymized metadata used to link microbiome samples to participants and the final feeding-group classification.

Columns:

| Column | Description |
|---|---|
| `Sample_ID` | Canonical study sample identifier |
| `QIIME2_Sample_ID` | Exact sample identifier present in the audited QIIME 2 feature tables and sample-level PICRUSt2 KO matrix |
| `Participant_ID` | Pseudonymized participant identifier |
| `Visit` | Study visit number |
| `Site` | Original study-specific oral sample-type code |
| `Study_group` | Final feeding-group classification used in the analyses: `Oral`, `Enteral`, or `Mixed` |

Audited structure:
- 221 unique samples
- 63 unique participants
- participant-level groups: Oral n = 29, Enteral n = 20, Mixed n = 14

#### Sample identifier mapping

For most samples, `Sample_ID` and `QIIME2_Sample_ID` are identical. For 25 samples, the identifiers differ only in zero-padding of the visit component. For example:

- `Sample_ID`: `s21.02.EPDNA`
- `QIIME2_Sample_ID`: `s21.2.EPDNA`

These identifiers refer to the same biological sample. The dedicated `QIIME2_Sample_ID` field allows exact joins to the original QIIME 2 artifacts and the sample-level KO matrix without modifying the original analytical files.

`Study_group` is the authoritative feeding-group variable for reuse of this curated dataset.

### Predicted functional data

#### `picrust2.sample.KO.relative.xls`

Sample-level PICRUSt2 matrix of predicted relative abundances of KEGG Orthologs.

Audited structure:
- 6,890 KEGG Orthologs
- 221 microbiome samples
- sample columns correspond exactly to `QIIME2_Sample_ID` in `study_metadata_public.csv`

This file is the sample-level KO layer retained for functional-data traceability.

#### `KEGG_Level3_participant_matrix.csv`

Participant-level matrix of predicted KEGG Level 3 pathway relative abundances used for the reported participant-level functional comparisons.

Audited structure:
- 221 KEGG Level 3 pathways
- 63 participants

Participant identifiers in this matrix are stored in lowercase form (for example, `s01`), whereas `Participant_ID` in `study_metadata_public.csv` is stored in uppercase form (for example, `S01`). The participant sets correspond exactly after case normalization.

Repeated sample-level information was summarized at the participant level before the reported KEGG Level 3 inferential comparisons.

## Analytical context

The associated manuscript reports:

- alpha diversity using Shannon and Simpson indices;
- beta diversity using Bray–Curtis dissimilarity with participant-level permutation procedures;
- differential taxonomic representation using LEfSe followed by participant-level sensitivity analyses;
- predicted functional profiling using PICRUSt2 and KEGG Level 3 pathways;
- participant-level pathway comparisons with correction for multiple testing.

PICRUSt2-derived values represent **predicted functional potential** inferred from marker-gene profiles and reference genomes. They are not direct measurements of microbial gene expression, protein abundance, pathway activity, or metabolic flux.

## Curation and provenance notes

1. Original QIIME 2 `.qza` artifacts are preserved rather than reconstructed for deposition so that their embedded provenance remains available.
2. The repository is intentionally minimal and does not attempt to reproduce every output generated during the outsourced sequencing/bioinformatics workflow.
3. Extracted `.qza`/`.qzv` directories, redundant visualizations, generic exploratory plots, alignment/tree outputs not required for the reported primary analyses, and duplicate open-format exports are not included.
4. The sample-level PICRUSt2 pathway files using `PWY-` identifiers are not included as substitutes for the KEGG Level 3 matrix because they represent a different pathway structure.
5. Absolute KO abundance outputs are not included because the manuscript reports relative predicted abundances.
6. Historical statistical outputs based on superseded grouping or analytical states are not part of this curated public package.
7. No retrospectively reconstructed analysis scripts are included solely for repository completeness.

## Recommended file linkage

For reuse:

1. use `study_metadata_public.csv` as the authoritative public metadata table;
2. use `QIIME2_Sample_ID` for exact joins to QIIME 2 sample identifiers and the sample-level KO matrix;
3. use `Participant_ID` to identify repeated samples from the same participant;
4. use `Study_group` as the final feeding-group variable;
5. normalize participant-ID case when joining `Participant_ID` to `KEGG_Level3_participant_matrix.csv`;
6. distinguish sample-level KO predictions from participant-level KEGG Level 3 pathway summaries.

## Reporting framework

Repository curation was guided by the STORMS reporting framework for human microbiome research and by the final audited analytical workflow of the study.

- STORMS: https://www.nature.com/articles/s41591-021-01552-x
- QIIME 2: https://qiime2.org/
- PICRUSt2: https://github.com/picrust/picrust2

## Citation

Please cite the Zenodo record:

**https://doi.org/10.5281/zenodo.19020777**

Once the associated article is published, users should cite both the article and the Zenodo dataset record.

## Contact

**Luanderson Lopes Pereira**  
Postgraduate Program in Dentistry and Health  
School of Dentistry, Federal University of Bahia  
Salvador, Bahia, Brazil  
Email: luannlopess2@gmail.com

## Version note

This README describes the curated repository package prepared after final analytical and repository audit in September 2026.