I wrote this script for a paper describing a rare disease caused by genetic variants of the CLN6 gene. The script served three purposes: 1. translating the Human Phenotype Ontology (HPO) term IDs of the patient records into HPO term names 2. replacing HPO terms of variable specificity with HPO terms at a defined, more general level 3. generating publication-ready plots of HPO term frequencies. Summarizing HPO terms at the organ system level revealed that the eye was the third most often affected organ in CLN6 disease. This was hidden by the heterogeneity of the original HPO annotations, and had not been noted in previous publications.

T-cells which recognise and kill cells expressing tumour-associated antigens are a promising new anti-cancer agent. However, in a few cases reactivity to antigens expressed by normal tissue caused serious adverse events. To mitigate such risks, Adaptimmune developed an extensive in-vitro testing pipeline. A pivotal element was predicting the repertoire of potential targets in human tissues that the therapeutic T-cells might respond to. The script reproduces two important figures of our paper. It involved checking > 10 million peptide sequences for matches in the human proteome; under 126 different criteria.

One of my favorite kind of plots for comparing data which might not be normally distributed.

The two figures together visualise a key message of the featured paper: T cells transduced with the AFPc332 T-cell receptor recognised target cells in vitro which expressed alpha fetoprotein (AFP) at levels found in ~30% of liver tumors. However, these levels were outside the range of AFP expression in normal tissues and diseased liver. The results supported the use of AFPc332 T-cells as a safe and effective anti-cancer agent.