Who this list is for
Biology titles often lead with a method. "Single-cell RNA sequencing", "CRISPR base editing" and "metagenomics" name tools, not questions, and a supervisor reading such a title still cannot tell what will be compared or measured.
How to use this list
Each table gives a dissertation title, a possible research question, a likely design and data source, and why the question matters now. Read the design column first. If you cannot name the material you will work with, the comparison you will make and the measurement that will answer the question, the title is not ready. Swap the species, tissue or site for one you can actually obtain, and check what your department can support before you commit. Neighbouring Academy lists are the biology topics on climate intervention, the network biology titles for computational work and the gene editing topics; all lists sit in the topics hub.
Theme 1: Cells, transcripts and molecular signatures
Single-cell and sequencing-based titles are attractive because the data are rich, and risky for the same reason. The analytical questions are settled enough to plan around: replication across biological samples, rather than sequencing depth, sets what a differential expression study can detect (Schurch et al., 2016). Microbiome titles need the same caution. Early taxonomic signatures of obesity were not reproduced consistently once compositional and confounding effects were examined (Finucane et al., 2014), although the functional work that followed (Turnbaugh et al., 2006) showed why the question remains open. A doctoral title in this theme should therefore name the comparison, the number of independent samples and the readout, not only the platform.
| Dissertation title | Possible research question | Likely design and data | Why it matters now |
|---|---|---|---|
| Single-cell transcriptional heterogeneity in one aggressive tumour type before and after first-line treatment | Which transcriptional cell states become more frequent in residual disease after treatment? | Paired patient samples; single-cell transcriptomes with per-patient replication | Treatment resistance is a property of cell populations, not of an average tumour |
| Uncultured gut taxa and chronic inflammatory disease: a case-control metagenomic study | Do people with the condition carry a different complement of uncultured taxa from matched controls? | Shotgun metagenomes with matched diet and medication covariates | Taxonomic associations need confounder control to be interpretable (Finucane et al., 2014) |
| Circular RNAs as candidate blood markers in Alzheimer's and Parkinson's disease | Do circular RNA levels differ between diagnosed cases and matched controls, and do they track severity? | Case-control cohort with a pre-registered marker panel and a validation set | Diagnosis still relies on clinical assessment and costly imaging |
Theme 2: Genome editing and model systems
Nuclease-based editing is guided to its target by an RNA molecule (Jinek et al., 2012), and guide RNA is not known to be imported into mitochondria, which is why mitochondrial editing uses protein-only deaminase editors rather than CRISPR tools (Mok et al., 2020). A title that promises CRISPR editing of mitochondrial DNA therefore misstates the problem. Model system titles need similar discipline about endpoints. Organoids are now used widely as experimental platforms (Liu et al., 2026), but their maturation is limited and claims about awareness or consciousness are not measurable, so the research question belongs with function and governance. Plant-made vaccine antigens have reached phase 3: a plant-derived particle vaccine showed 69.5% efficacy against symptomatic Covid-19 in a trial of 24,141 participants (Hager et al., 2022), which makes yield, stability and immunogenicity the live questions rather than feasibility in principle.
| Dissertation title | Possible research question | Likely design and data | Why it matters now |
|---|---|---|---|
| CRISPR-free base editing of mitochondrial DNA: editing outcomes and off-target profile in one cell model | What editing efficiency and off-target spectrum does a protein-only base editor achieve at defined mitochondrial sites? | Cell line panel; deep sequencing of edited and unedited positions, with heteroplasmy measured | Protein-only editors opened a target the guided tools cannot reach (Mok et al., 2020) |
| Functional maturation limits of iPSC-derived brain organoids and the governance gap | Which functional measures plateau as organoids age, and how do oversight frameworks treat those measures? | Longitudinal organoid cultures with electrophysiological readouts; documentary analysis of guidance | Oversight is being drafted while measurable endpoints are still contested (Liu et al., 2026) |
| Yield and immunogenicity of one plant-made vaccine antigen | How do expression host and antigen design affect yield and antibody response in a preclinical model? | Comparative expression study with immunisation in an animal model; standard antibody assays | Plant-made vaccines have shown efficacy in a phase 3 trial (Hager et al., 2022) |
Theme 3: Ecology, environment and whole organisms
Field and environmental titles stand or fall on sampling design. Phenology, the timing of seasonal events such as flowering and insect emergence, is one of the clearest cases: the question is whether two interacting species are shifting their timing at different rates, which requires repeated observation at the same sites across seasons rather than a single survey. Knowledge of these interactions is itself patchy, and Baracat et al. (2025) showed that combining distribution models for plants and their mutualists reveals large regions where interactions have never been recorded, which is a defensible way to justify a new field site. Microplastic and soil titles need the same narrowing: one short food chain or one crop and soil type, with an exposure measure and an outcome you can quantify.
| Dissertation title | Possible research question | Likely design and data | Why it matters now |
|---|---|---|---|
| Phenological mismatch between one flowering plant and its main pollinator along a temperature gradient | Are flowering and pollinator activity shifting at different rates, and does overlap shrink where warming is greater? | Repeated seasonal observation at paired sites; temperature records and overlap metrics | Interaction records are sparse for much of the world (Baracat et al., 2025) |
| Transfer of microplastics along one short aquatic food chain | How far do particle size and concentration explain burdens at the next trophic level? | Controlled feeding study or paired field sampling with particle counts per organism | Exposure is widespread while transfer rates remain poorly quantified |
| Designed microbial consortia and yield in one crop and soil type | Does inoculation with a defined consortium maintain yield at a reduced fertiliser rate? | Randomised pot and field trials with soil chemistry and yield as outcomes | Fertiliser costs and nutrient runoff are both under pressure |
Theme 4: Sleep, development and behaviour
That endpoint is not accessible in human participants, so the title has to split: either a human study with cognitive and actigraphy measures, or an animal model where neural measures are possible. Studies of adolescent wellbeing and behaviour also carry a measurement warning. Orben and Przybylski (2019) showed, using large national datasets, that associations between adolescent wellbeing and technology use were very small and highly sensitive to how the variables were specified, which is a reason to pre-register your analysis and report effect sizes rather than only significance.
| Dissertation title | Possible research question | Likely design and data | Why it matters now |
|---|---|---|---|
| Chronic sleep restriction and cognitive performance in adolescents: a within-participant study | How does cumulative sleep debt relate to attention and working memory across a school term? | Actigraphy with repeated cognitive testing; within-participant comparison | Small, specification-dependent effects require pre-registered analysis (Orben & Przybylski, 2019) |
| Sleep restriction and hippocampal measures in an adolescent rodent model | Does restricted sleep during adolescence change hippocampal measures and task learning? | Controlled animal study with a defined restriction protocol and behavioural testing | Neural endpoints that human studies cannot reach are testable in a model organism |
How to test a title before you commit
Apply the FINER criteria: is the study feasible, interesting, novel, ethical and relevant (Hulley et al., 2013)? For biology titles, check these in particular:
- Material: can you obtain the samples, cell lines, field sites or animals, with the approvals they require, within your registration period?
- Readout: is the thing you promise to measure measurable in that system, and by what assay?
- Replication: how many independent biological samples will you have, and what can that number detect (Schurch et al., 2016)?
- Controls: what is the comparison group, and how are confounders handled (Finucane et al., 2014)?
- Contribution: what will be known after your study that the published work cannot already show (Turnbaugh et al., 2006)?
For a fuller method see how to choose a PhD topic, how to select a PhD topic, topic selection in biology and our topic selection support.
Common mistakes
- Naming a platform instead of a question, so the title could describe a hundred different studies.
- Promising an endpoint the system cannot provide, such as consciousness in an organoid or synaptic plasticity in a human adolescent.
- Describing mitochondrial editing as a CRISPR application when the working tools are protein-only editors (Mok et al., 2020).
- Planning for sequencing depth and forgetting biological replication (Schurch et al., 2016).
- Treating a taxonomic or molecular association as a mechanism without a controlled follow-up.
- Choosing a field site or species for its interest value before checking access, permits and season.
Key takeaways
- A biology title should name the system, the comparison and the readout, not only the method.
- Mitochondrial base editing is CRISPR-free because guide RNA is not known to enter the organelle (Mok et al., 2020).
- Biological replication, more than sequencing depth, sets what a transcriptomic study can detect (Schurch et al., 2016).
- Microbiome and biomarker associations need matched controls and a validation set (Finucane et al., 2014).
- Test every title against feasibility, access to material and contribution before you commit (Hulley et al., 2013).
