Dataset DOI: 10.5061/dryad.1vhhmgr80

Description of the data and file structure

This repository hosts datasets describing metrics, time-series imaging-based latent representations, and gene expression derived from two-dimensional human gastruloids for the following manuscript in APL Bioengineering: Heterogeneous Patterning Behavior Emerges from Human Stem Cell-Based Embryo Model.

Files and variables

File: DAGOMR.zip

Endpoint Gastruloid Analysis

Features manually extracted from high-quality gastruloids at the final time point.

Endpoint Gastruloid Analysis.csv

Batch: experimental batch ID

Raft: microraft index

Channel: full (entire gastruloid) or SOX2 (SOX2+ region)

Feature: radius (radius in µm) or eccen (eccentricity)

Value: value of corresponding feature

Endpoint Gastruloid Analysis_Means.csv

Batch: experimental batch ID

Channel: full (entire gastruloid) or SOX2 (SOX2+ region)

Feature: radius (radius in µm) or eccen (eccentricity)

Mean: mean value of corresponding feature

Stdev: standard deviation of corresponding feature

Endpoint Gastruloid Grading

Quality grading of 50 fully developed gastruloids at the final time point.

Endpoint Gastruloid Grading.csv

Batch: experimental batch ID

Raft: microraft index

Grade: microraft grade [centered (centered gastruloid on microraft), defect (presence of microraft defect), aggregate (presence of adjacent aggregate), border (gastruloid contacting microraft border)]

SOX2_Center: whether gastruloid exhibits central circular SOX2+ region (yes or no)

Gastruloid Metadata

Metadata of gastruloid samples isolated for sequencing.

Gastruloid Metadata_RNA-seq Sample IDs.csv

Sample_Name: sequencing sample ID

Plate_Date: date of collection plate for sequencing

Well: well address

Array_Date: date of array (initial culture)

Batch: experimental batch ID

Raft: microraft index

Collect: whether gastruloid was successfully collected (0/1 for failure/success)

RNA: whether gastruloid was successfully sequenced (0/1 for failure/success)

Weight to Image Loss

16-dimension (16D) latent representations created from individual YX slices from time-series imaging of SOX2-mCitrine reporter in developing gastruloids while varying weight to image loss.

Img Loss_Lat Rep_Alpha{alpha value}.csv

dim_{n}: latent feature n value

timepoint: time point [0, 8]

sample_id: {array date}_array1_{raft number}

split: training (train), validation (val), or test (test) dataset

Img Loss_Lat Rep_Alpha{alpha value}_PCA.csv

sample_id: {array date}_array1_{raft number}

timepoint: time point [0, 8]

PC{n}: principal component n value

Latent Representations

16-dimensional (16D) latent representations created from individual YX slices (static) or TYX stacks (dynamic) from time-series imaging of SOX2-mCitrine reporter in developing gastruloids.

Lat Rep_Dynamic.csv

emb_{n}: latent feature n value

sample_id: {array date}_array1_{raft number}

split: training (train), validation (val), or test (test) dataset

Lat Rep_Dynamic_UMAP.csv

batch: experimental batch ID

raft: microraft index

umap_1: UMAP1 values

umap_2:  UMAP2 values

Lat Rep_Static.csv

emb_{n}: latent feature n value

sample_id: {array date}_array1_{raft number}

timepoint: time point [0, 8]

split: training (train), validation (val), or test (test) dataset

Lat Rep_Static_UMAP.csv

batch: experimental batch ID

raft: microraft index

timepoint: time point [0, 8]

umap_1: UMAP1 values

umap_2:  UMAP2 values

Model Loss as Function of Latent Dimensions

Model loss (image mean squared error and time point cross entropy) as a function of number of latent dimensions.

Loss Func of Lat Dim.csv

latent_dim: number of latent dimensions

image_loss_weight: weight to image loss [0, 1]

timepoint_loss_weight: weight to time point loss [0, 1]

best_test_image_loss: lowest image loss

best_test_image_epoch: minimum epochs to achieve lowest image loss

best_test_timepoint_loss: lowest time point loss

best_test_timepoint_epoch: minimum epochs to achieve lowest time point loss

Model Replicates

16-dimensional (16D) latent representations created from individual YX slices (static) from time-series imaging of SOX2-mCitrine reporter in developing gastruloids across multiple model runs/seeds.

Model Replicates_All Seeds_Displacement Vectors.csv

seed: model run seed [0, 4]

sample_id: {array date}_array1_{raft number}

timepoint: time point [0, 8]

dim_n: latent feature n

Model Replicates_Seed {seed number}_Static Lat Rep.csv

dim_n: latent feature n

timepoint: time point [0, 8]

sample_id: {array date}_array1_{raft number}

split: training (train), validation (val), or test (test) dataset

Model Replicates_Seed 3_UMAP_NNeighbors {number of neighbors}.csv

sample_id: {array date}_array1_{raft number}

timepoint: time point [0, 8]

umap_1: UMAP1 values

umap_2: UMAP2 values

Bulk RNA-seq

Sequencing data from single euploid and mosaic gastruloids.

Seq_Gene Expression Counts.csv

Gene: gene name

IJ.{plate date}.{well address}: gene expression count for sequencing sample ID

Seq_Gene Expression Counts_Normalized.csv

User_ID: gene name

ensembl_ID: gene name

symbol: gene name

search_label: gene name

IJ{plate date}{well address}: normalized gene expression count for sequencing sample ID

Seq_Heatmap_Clusters.csv

Cluster: gene expression cluster ID

Sample: sequencing sample ID

Seq_Heatmap_Data.csv

User_ID: gene name

ensembl_ID: ENSEMBL gene ID

symbol: gene name

search_label: {gene name} | {ENSEMBL gene ID}

IJ{plate date}{well address}: gene expression count normalized by standard deviation for sequencing sample ID

Seq_PCA.csv

Sample_ID: sample ID

PC{n}: principal component n value

Batch: experimental batch ID

Raft: microraft index

Seq_Sample IDs.csv

Experimental batch ID corresponding to each sequencing sample ID.

File: DAGOMR_Images.zip

Processed

Processed imaging time-series slices (YX) [{array date}_green_array1_{microraft index}_{time point}.png] of SOX2-mCitrine reporter in living gastruloids for each time point. Subdirectory names correspond to experimental batch ID.

Raw

Raw imaging time-series stacks (TCYX) [array1_raft{microraft index}_tl.tif}] of living gastruloids. Subdirectory “0623_Masks” contains manually drawn masks for the whole gastruloid (roi) and SOX2+ region (sox). Subdirectory names correspond to experimental batch ID.

Channels:

BF: brightfield (transmitted light)

SOX2: SOX2-mCitrine

Code/software

For code used, please refer to doi.org/10.5281/zenodo.17073719.