Phillipe Loher

Transfer RNA-derived fragments (tRFs) are noncoding RNAs that arise from either mature transfer RNAs (tRNAs) or their precursors. One important category of tRFs comprises the tRNA halves, which are generated through cleavage at the anticodon. A given tRNA typically gives rise to several co-expressed 5’-tRNA halves (5′-tRHs) that differ in the location of their 3′ ends. These 5′-tRHs, even though distinct, have traditionally been treated as indistinguishable from one another due to their near-identical sequences and lengths. We focused on co-expressed 5′-tRHs that arise from the same tRNA and systematically examined their exact sequences and abundances across 10 different human tissues. To this end, we manually curated and analyzed several hundred human RNA-seq datasets from NCBI’s Sequence Run Archive (SRA). We grouped datasets from the same tissue into their own collection and examined each group separately. We found that a given tRNA produces different groups of co-expressed 5′-tRHs in different tissues, different cell lines, and different diseases. Importantly, the co-expressed 5′-tRHs differ in their sequences, absolute abundances, and relative abundances, even among tRNAs with near-identical sequences from the same isodecoder or isoacceptor group. The findings suggest that co-expressed 5′-tRHs that are produced from the same tRNA or closely related tRNAs have distinct, context-dependent roles. Moreover, our analyses show that cell lines modeling the same tissue type and disease may not be interchangeable when it comes to experimenting with tRFs.

Motivation: We participated in the DREAM Single Cell Transcriptomics Challenge. The challenge’s focus was two-fold; a) to identify the top 60, 40 and 20 genes that contain the most spatial information, and b) to reconstruct the 3-D arrangement of the D. melanogaster embryo using information from those genes. Results: We developed two independent approaches, leveraging machine learning models from Lasso and Deep Neural Networks, that we successfully apply to high-dimensional single-cell sequencing data. Our methods allowed us to achieve top performance when compared to the ground truth. Among ~40 participating teams, the resulting solutions placed 10th, 6th, and 4th in the three DREAM sub-challenges #1, #2 and #3, respectively. Notably, for the Lasso approach we introduced a feature selection technique, Lasso-TopX, that allows a user to define a specific number of features they are interested in and the Neural Network approach utilizes weak supervision for linear regression to accommodate for uncertain or probabilistic training labels. Furthermore, we identified novel D. melanogaster genes that carry important positional information and were not previously suspected. Lastly, we show how the indirect use of the full datasets’ information can lead to data leakage and generate bias in overestimating the model’s performance. Availability: https://github.com/TJU-CMC-Org/SingleCell-DREAM/ . Contact: Nestoras.Karathanasis@jefferson.edu

We mined 10,274 datasets from The Cancer Genome Atlas (TCGA) for tRNA fragments (tRFs) that overlap nuclear and mitochondrial (MT) mature tRNAs. Across 32 cancer types, we identified 20,722 distinct tRFs, a third of which arise from MT tRNAs. Most of the fragments belong to the novel category of i-tRFs, i.e. they are wholly internal to the mature tRNAs. The abundances and cleavage patterns of the identified tRFs depend strongly on cancer type. Of note, in all 32 cancer types, we find that tRNA ^HisGTG produces multiple and abundant 5´-tRFs with a uracil at the -1 position, instead of the expected post-transcriptionally-added guanosine. Strikingly, these -1U His 5´tRFs are produced in ratios that remain constant across all analyzed normal and cancer samples, a property that makes tRNA ^HisGTG unique among all tRNAs. We also found numerous tRFs to be negatively correlated with many messenger RNAs (mRNAs) that belong primarily to four universal biological processes: transcription , cell adhesion , chromatin organization and development/morphogenesis . However, the identities of the mRNAs that belong to these processes and are negatively correlated with tRFs differ from cancer to cancer. Notably, the protein products of these mRNAs localize to specific cellular compartments, and do so in a cancer-dependent manner. Moreover, the genomic span of mRNAs that are negatively correlated with tRFs are enriched in multiple categories of repeat elements. Conversely, the genomic span of mRNAs that are positively correlated with tRFs are depleted in repeat elements. These findings suggest novel and far-reaching roles for tRFs and indicate their involvement in system-wide interconnections in the cell. All discovered tRFs from TCGA can be downloaded from https://cm.jefferson.edu/tcga-mintmap-profiles or studied interactively through the newly-designed version 2.0 of MINTbase at https://cm.jefferson.edu/MINTbase . NOTE: while the manuscript is under review, the content on the page https://cm.jefferson.edu/tcgamintmap-profiles is password protected and available only to Reviewers. Key Points: Complexity : tRNAs exhibit a complex fragmentation pattern into a multitude of tRFs that are conserved within the samples of a given cancer but differ across cancers. Very extensive mitochondrial contributions : the 22 tRNAs of the mitochondrion (MT) contribute 1/3 ^rd of all tRFs found across cancers, a disproportionately high number compared to the tRFs from the 610 nuclear tRNAs. Uridylated (not guanylated) 5´-His tRFs : in all human tissues analyzed, tRNA ^HisGTG produces many abundant modified 5´-tRFs with a U at their “-1” position (-1U 5´-tRFs), instead of a G. Likely central roles for tRNA ^HisGTG : the relative abundances of the -1U 5´-tRFs from tRNA ^HisGTG remain strikingly conserved across the 32 cancers, a property that makes tRNA ^HisGTG unique among all tRNAs and isoacceptors. Selective tRF-mRNA networks : tRFs are negatively correlated with mRNAs that differ characteristically from cancer to cancer. Mitochondrion-encoded tRFs are associated with nuclear proteins : in nearly all cancers, and in a cancer-specific manner, tRFs produced by the 22 mitochondrial tRNAs are negatively correlated with mRNAs whose protein products localize to the nucleus . tRFs are associated with membrane proteins : in all cancers, and in a cancer-specific manner, nucleus-encoded and MT-encoded tRFs are negatively correlated with mRNAs whose protein products localize to the cell’s membrane. tRFs are associated with secreted proteins : in all cancers, and in a cancer-specific manner, nucleusencoded and MT-encoded tRFs are negatively correlated with mRNAs whose protein products are secreted from the cell. tRFs are associated with numerous mRNAs through repeat elements : in all cancers, and in a cancerspecific manner, the genomic span of mRNAs that are negatively correlated with tRFs are enriched in specific categories of repeat elements. intra-cancer tRF networks can depend on sex and population origin : within a cancer, positive and negative tRF-tRF correlations can be modulated by patient attributes such as sex and population origin. web-enabled exploration of an “Atlas for tRFs” : we released a new version of MINTbase to provide users with the ability to study 26,531 tRFs compiled by mining 11,719 public datasets (TCGA and other sources).

Previously, we demonstrated that miRNA isoforms (isomiRs) are constitutive and their expression profiles depend on tissue, tissue state, and disease subtype. We have now extended our isomiR studies to The Cancer Genome Atlas (TCGA) repository. Specifically, we studied whether isomiR profiles can distinguish amongst the 32 cancers. We analyzed 10,271 datasets from 32 cancers and found 7,466 isomiRs from 807 miRNA hairpin-arms to be expressed above threshold. Using the top 20% most abundant isomiRs, we built a classifier that relied on “binary” isomiR profiles: isomiRs were simply represented as ‘present’ or ‘absent’ and, unlike previous methods, all knowledge about their expression levels was ignored. The classifier could label tumor samples with an average sensitivity of 93% and a False Discovery Rate of 3%. Notably, its ability to classify well persisted even when we reduced the set of used features (=isomiRs) by a factor of 10. A counterintuitive finding of our analysis is that the isomiRs and miRNA loci with the highest ability to classify tumors are not the ones that have been attracting the most research attention in the miRNA field. Our results provide a framework in which to study cancer-type-specific isomiRs and explore their potential uses as cancer biomarkers

For the anucleate platelet it has been unclear how well platelet transcriptomes correlate among different donors or across different RNA profiling platforms, and what the transcriptomes’ relationship is with the platelet proteome. We profiled the platelet transcriptome of 10 healthy young males (5 white and 5 black) with no notable clinical history using RNA sequencing and by Affymetrix microarray.

Human blood platelets are essential to maintaining normal hemostasis, and platelet dysfunction often causes bleeding or thrombosis. Estimates of genome-wide platelet RNA expression using microarrays have provided insights to the platelet transcriptome but were limited by the number of known transcripts. The goal of this effort was to deep-sequence RNA from leukocyte-depleted platelets to capture the complex profile of all expressed transcripts.