Skip to content

5. Conclusion and Future Work

Research conducted during this investigation has resulted in the identification of signature genes for two inflammatory skin diseases, AD and PSO, through the application of supervised ML algorithms. Several studies1,2,3 have stressed the heterogeneous nature of these diseases, making them a particularly challenging target for researchers seeking to identify signature genes. A full list of candidate signature genes for these diseases are discussed in Results. It has been previously shown that there are links between various forms of cancer and the skin diseases investigated in this project. Interestingly, many of the signature genes (for example SOX21-AS1, RP11-775C24.5 and C5orf46) identified in this project have also been linked to cancer. Further exploration of these genes could potentially form a strong basis for future work.

It is important to note that results from this study should be interpreted with caution due to the "black box" nature of the implemented ML algorithms4. It is particularly challenging, and in many cases not possible with current technology, to understand why the algorithms have selected certain genes above others. Although the results from this investigation appear promising, more experimental data is required on the identified signature genes to validate the efficacy of the findings. Furthermore, the study was limited to a single dataset comprised of 147 patients; any future work would do well to utilise a larger array of data, possibly including additional gene expressions and a higher number of samples.

Another limitation that should be acknowledged is that only the top 50 gene coefficeints for each condition were considered. Given the complex nature of the diseases such as AD, PSO and cancer, perhaps clusters of signature genes should be considered instead of individual genes, as the above mentioned diseases have several genes that are dysregulated at the same time. Future research may be focused on combining the already highlighted genes into clusters. Alternatively, higher weightings could be allocated artificially to genes that have already been identified in the literature as playing an important role for AD and PSO at the models' training stage.

To conclude, this study has achieved four Research Objectives to enable addressing the Research Aim of identifying AD- and PSO-related genes from high-dimensional RNA-Seq data.


I would like to thank my supervisor, Dr Reiko Tanaka, for her support and guidance throughout the project. I would also like to thank Post-Doctoral Research Associate, Sun Yuxin, for her helpful discussions.

  1. Yanoff M, Sassani JW. Skin and Lacrimal Drainage System. In: Ocular Pathology. Elsevier; 2020. p. 163-233.e10. 

  2. Lowes MA, Russell CB, Martin DA, Towne JE, Krueger JG. The IL-23/T17 pathogenic axis in psoriasis is amplified by keratinocyte responses. Vol. 34, Trends in Immunology. Elsevier Current Trends; 2013. p. 174–81. 

  3. Noda S, Krueger JG, Guttman-Yassky E. The translational revolution and use of biologics in patients with inflammatory skin diseases. J Allergy Clin Immunol. 2015 Feb 1;135(2):324–36. 

  4. Castelvecchi D. Can we open the black box of AI? Nature. 2016 Oct 5;538(7623):20–3.