Regulation of transcript isoforms

Theme: World-Class Underpinning Bioscience

Aims:
1. Benchmark the performance of isoform quantification tools run on single-cell RNA- seq data.
2. Characterise the impact of technical noise on isoform quantification in single-cell RNA-seq data.
3. Determine how the patterns of isoform expression seen in populations of cells are generated by individual cells.
4. Consider how patterns of isoform expression are regulated.

A brief explanation of these aims is given as follows. Use of single-cell RNA-seq is becoming increasingly popular despite a lack of knowledge regarding best bioinformatic practices for analysing the data. Single-cell RNA-seq could be particularly informative regarding alternative splicing. However, before this can be fully addressed, data processing issues need to be explored. One aim of this PhD project is to benchmark the performance of existing isoform quantification tools when run on single cell data. This will establish whether existing isoform quantification tools are capable of accurately performing isoform quantification when applied to single-cell data and will determine which tool performs best. In addition, the impact of technical noise on isoform quantification must be considered.

Drop-outs are more common for lowly expressed genes (Kharchenko et al., 2014), consequently the expression of lowly expressed genes and isoforms is less likely to be detected in single cells. It will be important to determine whether there is a threshold level of expression below which performing isoform quantification and splicing analysis is not meaningful for single-cell RNA-seq data, and if so where that threshold lies.

With this information, it will be more meaningful to use single-cell RNA-seq data to address biological questions as the limitations of current single-cell RNA-seq technologies will be more fully understood. The next aim will be to determine how the patterns of isoform expression seen in populations of cells are generated by individual cells - whether each cell expresses all the isoforms seen at the population level, or whether individual cells express only a subset of the isoforms seen at the population level. A comprehensive understanding of drop-outs will be important when answering this question, as a high rate of drop-outs could give the impression that less isoforms are expressed in individual cells than is truly the case.

It is plausible based on current evidence that there is some degree of heterogeneity in the number of isoforms expressed in the individual cells making up a population (Marinov et al., 2014; Shalek et al., 2013; Zhao et al., 2016). Assuming this is the case, the next aim will be to determine how the number of isoforms expressed in a single cell is regulated. An obvious candidate for how this regulation could be effected is epigenetic marks. Machine learning approaches could be used to investigating this possibility.

It is possible that the number of isoforms expressed in each cell in a population is largely homogenous between cells, or that it is not possible to distinguish whether the heterogeneity present is biological or technical in origin. If this proves to be the case, the regulation of the number of isoforms expressed in populations of cells could instead be considered. Again, epigenetic marks are a clear candidate for how the number of expressed isoforms might be regulated.