https://doi.org/10.5061/dryad.ghx3ffbwj

ABIS is a prospective, general population-based cohort study to which parents of all children born in southeast Sweden during the period 1 October 1997–1 October 1999 were invited. Of the 21,700 families asked, 17,055 agreed to participate (78.6%), giving informed consent after receiving oral and written information. The children have been followed from birth. The ABIS study was approved by the Research Ethics Committee, Linköping University and Lund University, approval numbers Dnr 03-092; LiU 287.96, LU 83-97, LiU 321-99.and Linköping University Dnr 36287, Dnr 03-513, Dnr 2018/380-32.

Description of the data and file structure

The data analyzed in this investigation involve ABIS Questionnaires (at birth, one year, three years, and five years of age), microbiome (16S rRNA sequencing on Illumina MiSeq V3-V4 on stool samples collected at one year of age), and metabolomic datasets (cord serum and stool collected at one year of age). Processed datasets corresponding to these datasets are provided, along with descriptions, below. Results and the corresponding analytical methods employed to capture these differences are elaborated upon in our accompanying manuscript published in Cell.

1- ABIS Questionnaires & HLA

This investigation focuses on analyzing early-life factors and genetic markers associated with future ND diagnoses in ABIS. The earliest questionnaires, obtained at ages one, three, and five, were analyzed, and the latest diagnoses were obtained when the children were 21 to 23 years old. The cohort consists of 14,869 controls and 1,179 children with a future ND diagnosis as of December 2022. Human leukocyte antigen (HLA) class II genotype was determined for 3,783 children using sequence-specific hybridization. Given the known association of autoimmune disease with neurodevelopment, this investigation also compared the prevalence of risk alleles for autoimmunity across future NDs and controls.

File Descriptions

The cohort is represented across five CSV files, each containing data from controls and children with future diagnosis of the following:

• 1.1- ND_DRYAD_v2.csv contains data from controls and all children with a future ND diagnosis.
• 1.2- ADHD_DRYAD_v2.csv contains data from controls and all children with a future ADHD diagnosis.
• 1.3- ASD_DRYAD_v2.csv contains data from controls and all children with a future ASD diagnosis.
• 1.4- IntDis_DRYAD_v2.csv contains data from controls and all children with a future intellectual disability diagnosis.
• 1.5- SpeechDis_DRYAD_v2.csv contains data from controls and all children with a future speech disorder diagnosis.

Data in each file represents a combination of factors based on their significance with the respective outcome:

2- Infant Microbiome Datasets & Files

Extraction of DNA from fecal samples collected at one year of age, followed by amplification of the V3-V4 region using 16S rRNA-PCR, was performed. A total of 1,748 samples underwent sequencing in ten pools through Illumina MiSeq 2x300 bp at the Interdisciplinary Center for Biotechnology Research (ICBR) at the University of Florida, Gainesville, Florida, USA. Amplicons for targeted V3-V4 16S rRNA sequencing were generated using standard Illumina Read 1 sequencing/indexing primers: 341F (NNNNCCTACGGGAGGCAGCAG) and 806R (GGGGACTACVSGGGTATCTAAT). Exclusion criteria led to the removal of five samples exhibiting markedly low reads. Sample inference employed the dada function with multithreading, and sequence tables from the ten pools were amalgamated prior to applying the consensus-method chimera removal. Environmental factor analysis considered all ABIS samples, excluding those with low counts (n=1,743).

• 2.1- Factors_Data_Dryad.csv contains factors associated with increased risk or protection from NDs, along with raw counts for the 500 most prevalent amplicon sequence variants (ASVs). These factors were investigated to understand their correlation with microbial taxa at one year of age. The examined factors include experiences such as gastroenteritis, antibiotic treatment, and ear infections during the first year of life, psychosocial vulnerability, maternal smoking during pregnancy, durations of total/exclusive breastfeeding, consumption frequency of chocolate, fries, and chips in the first year, and mode of delivery. All factors underwent dichotomization for subsequent differential analysis. Breastfeeding durations (total/exclusive) were categorized into one to four months versus five or more months, while psychosocial vulnerability was dichotomized into high and low risk based on the total index.

• 2.2- Sym_Data_Dryad.csv contains symptom data at 3 and 5 years, as well as the raw counts for the 500 most prevalent ASVs. At the child’s three and five years of age, parents involved in the ABIS study provided responses to 11-12 binary-response questions related to growth, mood, and gastrointestinal issues on the questionnaire. They answered the following question: ‘Do you think that the child suffers from or is affected by…’. Subsequent symptoms encompassed poor growth, poor weight gain, poor appetite, stomachache, bloated/gassy stomach, diarrhea three times or more per day, vomiting three times or more per day, constipation, fatigue, general irritation, cranky mood/screaming, and poor sleep quality. Presence of these symptoms, as reported by the parent on the ABIS questionnaire, is marked with a “1” in the dataset.  Significant differences in the accompanying manuscript in Cell were determined using DESeq2 with FDR correction, on all taxa at the ASV and species level that passed filtering thresholds.

  Symptoms reflected in the 3-year and 5-year questionnaires are coded as the following. The letters preceding each symptom in the variable name are also used in the symptom clusters, defined in the section below:

  Symptoms at 3 years	Symptoms at 5 years
  a_growth	a_growth.5
  b_weight	b_weight.5
  c_appetite	c_stomachache.5
  d_stomachache	d_appetite.5
  e_bloated	e_gassy.5
  f_diarrhea	f_diarrhea.5
  g_vomiting	g_vomiting.5
  h_constipation	h_constipation.5
  i_fatigue	i_fatigue.5
  j_irritation	j_irritation.5
  k_cranky	k_sleep.5
  l_sleepquality

  • Calculated Sums and Cumulative Symptom Variables:
    • The sum of the 12 symptoms on the 3-year questionnaire was calculated. Subsequently, the sums of symptoms corresponding to principal components analysis are denoted as [c16_sum_agijkl], [c16_sum_deh], [c_16abc], where each letter corresponds to the symptom on the 3-year questionnaire.
    • The sum of the 11 symptoms on the 5-year questionnaire was also calculated. Subsequently, the sums of symptoms corresponding to principal components analysis are denoted as [e_14ijk], [e_14abd], [e_14ce], where each letter corresponds to the symptom on the 5-year questionnaire.
    • Cumulative symptoms were dichotomized as “0” or “1” based on whether the child had any of the symptoms in that symptom cluster.
      Dichotomization of symptoms clusters in this data file is defined as follows:
      • Three years of age
        • c16_sum_1plus: 0= no symptoms reported by the parent at 3 years; 1= one or more symptoms
        • c16_sum_agijkl_1plus: 0= no symptoms in the agijkl symptom cluster; 1= one or more symptoms in the agijkl symptom cluster
        • c16_sum_deh_1plus: 0= no symptoms in the deg symptom cluster; 1= one or more symptoms in the deg symptom cluster
        • c16_sum_abc_1plus: 0= no symptoms in the abc symptom cluster; 1= one or more symptoms in the abc symptom cluster
      • Five years of age
        • e_sum_1plus: 0= no symptoms reported by the parent at 5 years; 1= one or more symptoms
        • e_14ijk_1plus: 0= no symptoms in the ijkl symptom cluster; 1= one or more symptoms in the ijkl symptom cluster
        • e_14abd_1plus: 0= no symptoms in the abd symptom cluster; 1= one or more symptoms in the abd symptom cluster
        • e_14ce_1plus: 0= no symptoms in the ce symptom cluster; 1= one or more symptoms in the ce symptom cluster
• 2.3- Microbiome_ASD_earlylatediagnosis_Dryad.csv contains the designations for early and late ASD diagnoses for the 39 ABIS children included in the infant microbiome analysis, along with their biological sex and the relative abundances of the 500 most prevalent ASVs. Differential analysis was conducted separately for males and females due to variations in the average age of diagnosis (13.4 ±4.0 years for males; 17.6 ±2.6 years for females). The distinction between early and late diagnoses was based on the mean age of diagnosis, stratified by biological sex.
• 2.4- Microbiome_NDOutcomes_Dryad_short.csv contains the future neurodevelopment (ND) outcomes (ASD, Intellectual Disability, Speech Disorder, or ADHD) for ABIS children with 16S rRNA data at one year of age, as well as the relative abundances of the 500 most prevalent ASVs. Diagnoses up until Dec 2020 were considered in this manuscript. The sum of future ND diagnoses per child [Sum_ND] is provided, while [NDev] demonstrates whether the child was considered a future case or control. Significant differences in the manuscript were determined using DESeq2 with FDR correction, on all taxa at the ASV and species level that passed filtering thresholds.
• 2.5- TopGen_StoolNegativeMode_Dryad.csv contains the relative abundances of Bifidobacterium, Coprococcus, Roseburia, Akkermansia, and Faecalibacterium at one year of age. This dataset also includes peak heights derived from untargeted mass spectrometry analysis of fecal samples from individuals with future ASD (n=23) and their matched controls (n=23). In the negative ionization mode, a total of 300 known metabolites were identified. Employing random forest regressors in Python, the top twenty stool metabolites in the negative-ion mode that most effectively predicted the relative abundance of Bifidobacterium, Roseburia, Faecalibacterium, Akkermansia, or Coprococcus were determined.

3- Infant Stool Metabolomics Datasets & Files

Untargeted LC-MS metabolomic analysis was conducted on 23 individuals with future ASD diagnosis and 23 controls, matched by propensity score on biological sex and municipality, and with no difference in child age at stool collection. Stool samples underwent cellular extraction and pre-normalization before comprehensive metabolomics profiling at the Southeast Center for Integrated Metabolomics (SECIM) at the University of Florida, Gainesville, FL. A Thermo Q-Exactive Oribtrap mass spectrometer, coupled with Dionex UHPLC and autosampler, was used for both positive and negative heated electrospray ionization, achieving a mass resolution of 35,000 at m/z 200 in separate injections. Metabolomic features were cross-referenced with SECIM’s internal retention time metabolite library (containing 1,414 compounds) for identification based on established metabolomic databases. Fatty acids were quantified using LipidMatch.

• 3.1, 3.2- NegIonStool_CodeBook_Dryad.csv and PosIonStool_CodeBook_Dryad.csv contain the mass to charge ratio (m/z), retention time, molecular formula, and identification for each stool metabolite identified by the negative and position ionization modes, respectively.
• 3.3, 3.4- Untargeted_Negative_Dryad.csv and Untargeted_Positive_Dryad.csv contain the peak heights resulting from untargeted mass spectrometry analysis of future ASD (n=23) and control samples (n=23), representing a total of 300 metabolites identified in the negative ionization mode and 271 in the positive ionization mode.
• 3.5- equol_Dryad.csv contains the peak heights of equol (m/z 241.0866), by future ASD diagnosis [ASD_20].
• 3.6- StoolMetabolites_Fattyacids_Dryad.csv contains the sum of all peaks [peak_Sum], total moles of 16.1 [moles_16.1], and molar percentages of the 22 fatty acids that were identified, by future ASD outcome [ASD_20].
• 3.7- StoolMetabolites_SCFAs_Dryad.csv contains the peak heights identified for butyrate and citrate, based on future ASD status [ND].

4- Cord Serum Metabolomics Dataset & Files

Cord serum samples were obtained from maternal donations. Samples were analyzed using an ultra-high-performance liquid chromatography quadrupole time-of-flight mass spectrometry (UHPLC-QTOFMS from Agilent Technologies; Santa Clara, CA, USA). In this investigation, lipidome and metabolome differences were investigated after controlling for various prenatal risk factors, education of the parents, and gestational age using propensity score matching to select matched controls (n=27) for future ASD cases (n=27). Correlation analysis between

• 4.1- Identification_CordBloodMetabolites_Dryad.csv contains the m/z and retention time for each of the known polar metabolites in cord serum.
• 4.2- CordBlood_ProMatch_Metadata_Dryad.csv contains the maternal factors from the birth questionnaire, upon which samples were propensity-score balanced and selected for differential analysis, along with the peak heights from the 112 known polar metabolites identified in cord serum. The variables included are described below.

• 4.3- CordBlood_Microbiome_Dryad.csv contains the relative abundances of 13 microbial species and seven ASVs that were consistently associated with NDs in our investigation, along with seven polar cord serum metabolites and 16 lipids of interest due to the identified association with NDs in the ABIS material. Correlation analysis between the relative abundances and these selected features in the cord serum metabolome and lipidome were carried out.

Sharing/Access information

Any additional information required to reanalyze the data reported in our paper is available from the lead contact upon request and an appropriate institutional collaboration agreement. Indirect identifiers were kept at 3 maximum within each dataset to satisfy Dryad human subjects data requirements.