25Single cell RNA-seq (scRNA-seq) over specified time periods has been widely 26 used to dissect the cell populations during mammalian embryogenesis. 27Integrating such scRNA-seq data from different developmental stages and from 28 different laboratories is critical to comprehensively define and understand the 29 molecular dynamics and systematically reconstruct the lineage trajectories. Here, 30we describe a novel algorithm to integrate heterogenous temporal scRNA-seq 31 datasets and to preserve the global developmental trajectories. We applied this 32 algorithm and approach to integrate 3,387 single cells from seven heterogenous 33 temporal scRNA-seq datasets, and reconstructed the cell atlas of early mouse 34 cardiovascular development from E6.5 to E9.5. Using this integrated atlas, we 35 identified an Etv2 downstream target, Ebf1, as an important transcription factor 36 for mouse endothelial development. 37 38 function based on stochastic neighbor assignment in the transformed space, 129 similar to the neighborhood component analysis (NCA) [23]. It should be noted 130 that although the CLN was evaluated for an individual time point, a single linear 131 transformation applies to the cells from all time points. Thus, unlike our previous 132 work on visualizing temporal scRNA-seq data[10], the cells from different time 133 points will not be confined to the same transformed (low) dimensional space. 134 135