When I started contributing to BugSigDB, I knew I would be working with scientific literature, extracting information, and organizing microbial findings into a structured database.
What I did not fully appreciate was how much interpretation this work would require.
I had a background in Computer Science, with a specialization in Data Science, and had previously worked on academic research during my master's degree. I was comfortable reading scientific papers, understanding research methods, and working with data.
But microbiome study curation introduced a different kind of challenge.
It required me to pay attention to details I might otherwise have overlooked: the exact groups being compared, the direction of microbial abundance, the statistical methods used, the taxonomic level of reported findings, and the terminology used to describe health conditions.
Over the course of my first contributions, I made mistakes, received feedback, revisited papers, and gradually became more confident in my decisions.
In this article, I want to share some of those experiences. Not as a perfect record of everything I did, but as an honest reflection on what I learned while contributing to BugSigDB.
1. My First Contribution: The Mini-Curation
My first task was the introductory mini-curation provided through GitHub issue #94 – Welcome Outreachy Applicants!.
The exercise involved comparing the gut microbiome of individuals living with HIV against individuals without HIV.
I had to identify the relevant study information, understand the comparison, and determine which microbial taxa were less abundant in the HIV group.
One of the first things I encountered was an ontology issue.
The HIV-related term I initially came across, EFO:0000764, was obsolete. I had to investigate the appropriate replacement, which led me to MONDO:0005109.
This was a small but important introduction to the role of ontologies in BugSigDB.
I learned that the terminology used in a scientific paper does not always correspond directly to the terminology available in a database. Sometimes, a term has been deprecated or replaced, and you need to investigate before selecting an alternative.
The exercise also reinforced something that became important throughout my subsequent contributions: understanding the comparison is essential to recording microbial abundance correctly.
My mini-curation was accepted, and I moved on to working with full studies.
2. My First Full Study: Learning to Read Beyond the Abstract
My first full study was Effect of prophylactic dextrose gel on the neonatal gut microbiome.
- GitHub issue: #1928
- PMID: 34857640
- DOI: 10.1136/archdischild-2021-322757
- View my BugSigDB curation
This study examined the effect of prophylactic dextrose gel on the neonatal gut microbiome. It also investigated other factors associated with differences in the neonatal microbiome, including mode of delivery and feeding practices.
I worked on three experiments.
The first compared infants delivered by caesarean section with those delivered vaginally. The second examined feeding groups, including formula-fed infants with or without breastmilk compared with infants receiving breastmilk only. The last compared Indians with other ethnic groups
The mistake: Getting abundance directions mixed up
One of the important corrections I had to make involved the direction of microbial abundance.
For the delivery-mode comparison, I initially mixed up the increased and decreased taxa.
The relevant findings were reported in Supplementary Material 4, which contained the results of the MaAsLin2 analysis.
There were taxa that were more abundant in infants delivered by caesarean section and others that were less abundant.
I had to revisit the supplementary material and correct the direction of the signatures.
This taught me not to rely on my initial interpretation of a table or figure. I needed to verify exactly which group had the higher or lower abundance.
Another lesson: Understanding study design
The study was described as an observational cohort study nested within a randomized trial.
That distinction mattered.
Although the larger trial involved randomization, the microbiome study I was curating was an observational cohort analysis. I had to understand the actual design of the analysis rather than simply classify the paper as a randomized controlled trial because the word "randomized" appeared in its description.
Feedback from my mentor (Chloe) reinforced the importance of choosing the best-fitting study design rather than selecting multiple descriptions just because they seemed relevant.
What I learned: A paper can have a complicated research structure. You need to understand the particular analysis being curated, not just the broader study it belongs to.
3. Working with Figures: The Vaginal Microbiome and Water-Based Lubricants
My next interesting study was The In Vivo Effect of Water-Based Lubricants on the Vaginal Microbiome of Women from Varying Age Groups.
- GitHub issue: #2077
- PMID: 39338590
- DOI: 10.3390/microorganisms12091917
- View my BugSigDB curation
I found this study particularly interesting because it examined the vaginal microbiome in relation to women's age groups and menopausal status, alongside an investigation of water-based lubricants.
For my curation, I worked with two comparisons:
- Postmenopausal women versus premenopausal women.
- Older women versus younger women.
I used the relevant findings in Figures 3C and 3D to identify the microbial signatures.
For the menopausal comparison, for example, the results included genera that were more abundant in postmenopausal women and Lactobacillus, which was less abundant in that group.
This study helped me become more comfortable extracting microbial findings directly from figures rather than relying exclusively on the written Results section.
It also reminded me that a study's overall objective does not automatically tell you which findings belong in a particular experiment. You must identify the actual comparisons and the evidence supporting each one.
What I learned: Figures are not merely illustrations accompanying a scientific paper. They can contain essential information for curation, and learning how to interpret them is an important part of the process.
4. Learning to Interpret Statistical Findings: A Study on Menopause
Another study I worked on was Association of Vaginal Microbiota With Signs and Symptoms of the Genitourinary Syndrome of Menopause Across Reproductive Stages.
- GitHub issue: #1931
- PMID: 33903897
- DOI: 10.1093/gerona/glab120
- View my BugSigDB curation
This study investigated associations between vaginal microbiota and symptoms of genitourinary syndrome of menopause across different reproductive stages.
One of the things I had to pay attention to was how the condition was represented.
I initially used a more specific phrase, "postmenopausal atrophic vaginitis," but changed the condition to "atrophic vaginitis" after reconsidering the terminology and the appropriate representation.
This was another reminder that the condition being studied and the wording used in the paper need to be interpreted carefully.
I also spent time understanding odds ratios, confidence intervals, and p-values.
For example, when interpreting an odds ratio, the confidence interval matters. A confidence interval that excludes 1 indicates a statistically significant association at the corresponding confidence level, assuming the usual interpretation of the interval.
I learned to pay attention not just to whether an association was reported, but also to its direction and statistical support.
What I learned: Statistical results should not be copied without interpretation. You need to understand what the reported measure represents and what conclusions it supports.
5. My First Collaboration: Curating a Study with Another Contributor
One of my most valuable experiences was my first collaborative curation.
The paper was The cervical microbiota of Hispanics living in Puerto Rico is nonoptimal regardless of HPV status.
- PMID: 37534938
- DOI: 10.1128/msystems.00357-23
- View the BugSigDB study and collaborative curation
I worked on the first three experiments, while another contributor, Sally Kinyua, worked on the remaining experiments.
This was my first experience dividing the curation of a single scientific paper with another contributor.
The study was particularly interesting because it investigated the cervical microbiota across different physiological stages and characteristics, including pregnancy, menopause, HPV status, and cervical health.
6. A Study That Made Me Look Beyond the Microbiome
Another contribution that stood out to me was Characterization and correlation of gut microbiota in major depressive disorder patients with suicidal ideation.
- PMID: 42768327
- DOI: 10.1186/s12888-026-08290-9
- View my BugSigDB curation
I was particularly interested in this paper because of its subject matter. It investigated gut microbiota in people with major depressive disorder, comparing participants with and without suicidal ideation, alongside healthy controls.
The study involved several analyses, including differential-abundance testing and correlation analyses.
This meant I had to be careful about distinguishing microbial taxa identified through differential-abundance analysis from taxa discussed in relation to correlations with clinical measurements.
For example, a bacterium being correlated with a depression score does not automatically mean that it was significantly more or less abundant in a particular comparison group.
These are different types of findings, and they should not be represented as though they mean the same thing.
The study also involved several comparisons:
- Major depressive disorder with suicidal ideation versus healthy controls.
- Major depressive disorder with suicidal ideation versus major depressive disorder without suicidal ideation.
- Major depressive disorder without suicidal ideation versus healthy controls.
Working through these comparisons reinforced the importance of keeping the experimental groups and their definitions clear.
What I learned: A scientific paper can report several interesting relationships, but a curator must understand which results represent differential abundance and which represent other statistical associations.
7. The Biggest Lessons from My First Contributions
Looking back at these experiences, there are several lessons I keep returning to.
Read the supplementary materials
Some of the most important information I needed was not in the main text.
Supplementary tables helped me identify microbial taxa, verify abundance directions, and understand the statistical analyses.
Do not stop at the abstract or even the Results section. Check the supplementary materials whenever the paper refers to them.
Understand the comparison before recording the signature
Always know which group is Group 0 and which is Group 1.
A signature is meaningful only when its direction is correctly associated with the comparison.
Do not confuse different types of statistical findings
Differential abundance, alpha diversity, correlations, and associations are not interchangeable.
A paper may report all of them, but each answers a different question.
Pay attention to taxonomic resolution
Microbial names can be reported at different taxonomic levels.
A finding reported at the genus level should not be casually converted into a species-level entry. Similarly, an ASV or an unclassified taxon requires careful interpretation.
Use the appropriate taxonomy resources and follow the Curation Policy.
Feedback is part of the contribution process
Receiving corrections does not mean you have failed.
In fact, some of my most useful lessons came from reviewing feedback, revisiting the paper, and understanding why a particular entry needed to change.
The goal is not to defend every decision you make. It is to produce an accurate curation.
Keep track of your decisions
As you work through a paper, record where you found important information.
Was it in Figure 2? Supplementary Table 4? The Methods section?

Figure 1: A screenshot from my study notes with several important facts highlighted
Keeping track of these sources makes it easier to verify your work and explain your decisions during review.
8. Final Thoughts
My first BugSigDB contributions taught me that scientific data curation is much more than transferring information from a paper into a database.
It requires reading carefully, interpreting research methods, understanding statistical results, identifying microbial taxa, and making decisions that preserve the meaning of the original research.
I made mistakes. I received feedback. I corrected my work. I also had the opportunity to collaborate with another contributor and learn from the experience.
What I appreciate most is how each study introduced me to a slightly different challenge. One taught me to pay closer attention to abundance directions. Another made me more comfortable interpreting figures. Others helped me understand study design, statistical findings, taxonomy, and collaboration.
I came into this project with a Data Science background, but I am leaving these early contributions with a much better appreciation of the work involved in making scientific findings structured, searchable, and reusable.
And perhaps the biggest lesson is this:
You do not have to get everything right on your first attempt. You do have to be willing to investigate, learn from feedback, and correct your work.
That is how you become a better curator, one study at a time.
Happy curating!
Top comments (0)