Day 1: Introduction to computational precision medicine and working
with expression data in R
When: Tuesday, 1 September 2026, 13:00–17:00
What: Lecture, Q&A, exercises, and self-study
Program
13.00 - 14.00: Lecture: introduction to precision medicine +
“theoretical brush up”
14.00 - 14.15: Break
14.15 - 17.00:
Setting up RStudio; Exercise: working with gene expression data in R;
Self Study
Today’s learning objectives
After today’s session and going through the reading materials, you
will be able to:
- Explain the basic concepts of precision medicine and computational
precision medicine.
- Describe how bioinformatics can contribute to precision diagnostics,
prognostics, and therapeutics.
- Understand the difference between the data coming out of RNA-seq
experiments and microarrays.
- Distinguish between raw counts, normalization, and
transformation.
- Explain why RNA-seq expression data are normalized and
transformed.
- Understand and apply principal component analysis (PCA) on
gene-expression data.
- Distinguish biological variation from potential technical variation
such as batch effects.
Note: - when we write “understand”, we mean getting the basic
intuition of concepts, and you will not be asked about the underlying
biology or math, unless otherwise specified.
- whenever we write
“apply”, it is implied that this is in R.
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to know
about these topics. It is not a comprehensive list of what you need to
know!
1. What can PCA be used for?
a) Visualization of
multidimensional sample differences
b) Dimensionality reduction
c) Visualization of batch effects
d) All of the above
2.
What do the x- and y-axis coordinates in a PCA scores plot
represent?
a) Original gene-expression values
b) Percentage of
variance explained
c) Eigenvalues
d) Principal component
scores
3. What are the units of gene quantification in
RNA-sequencing?
a) Transcripts per million
b) Fragments per
kilobase per million
c) Number of reads mapped to a given
transcript
d) All of the above
4. Which probability
distribution is commonly used by DESeq2 to model RNA-seq count data?
a) Normal distribution
b) Beta distribution
c) Negative binomial
distribution
d) T distribution
5. What is the purpose of
computational precision medicine?
a) Finding and evaluating new
therapeutic targets
b) Supporting the practices of precision
medicine with data analysis
c) Supporting of precision diagnostics
by large-scale data analysis
d) All of the above
6. What
are batch effects?
a) Biological variance that we want to remove
b) Biological variance that we want to keep
c) Technical variance
that we want to remove
d) Technical variance that we want to
keep
7. Which statement about the number of principal
components produced by PCA is correct?
a) PCA always produces two
principal components
b) PCA always produces the same number of
components as there are samples
c) For centered data with \(n\) samples and \(p\) features, there can be at most
min(\(p\), \(n-1\)) non-zero principal components
d)
PCA always produces one fewer component than there are features
8. What is a primary purpose of the UCSC Xena platform?
a) To
interactively explore and visualize genomic and clinical datasets
b)
To sequence RNA from biological samples
c) To identify mutations
directly from raw sequencing reads
d) To perform laboratory
experiments on tumor samples
9. Which of the following are NOT
a hallmark of cancer?
a) Replicative immortality
b) Resistance
to immunotherapy
c) Resistance to programmed cell death
d)
Genomic instability
Self-evaluation quiz answers (click to expand)
d
d
d
c
d
c
c
a
b
Curriculum (click to expand)
Short history
of gene expression profiling
Next generation
sequencing of RNA (RNA-seq)
Introduction to
RNA-seq data
The Cancer Genome
Atlas Legacy: Pushing the Boundaries of Research
Introduction
to the UCSC Xena data collection
Introduction
to the molecular pathogenesis of cancer (also known as “the hallmarks of
cancer”)
Your turn! Find resources for these concepts and how to
do them in R:
* Batch effects/technical variance in gene expression
data
* Principal component analysis
Day 2: Differential expression analysis and gene set enrichment
analysis
When: Tuesday, 8 September 2026, 13:00–17:00
What: Lecture, Q&A, exercises, and self-study
Program
13.00 - 13.45: Lecture: Differential expression analysis and gene set
enrichment analysis
13.45 - 14.00: Break
14.00 - 17.00:
Exercise: Differential expression analysis and gene set enrichment
analysis; Self-study
Today’s learning objectives
After today’s session and going through the reading materials, you
will be able to:
- Explain the purpose of differential expression analysis.
- Perform a basic differential expression analysis of RNA-seq
data.
- Understand the rationale behind gene set enrichment
analysis.
- Interpret gene-expression changes in a biological context.
- Understand the need for multiple-testing correction and interpret
adjusted p-values (FDR).
- Apply and interpret volcano plots.
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to
know about these topics. It is not a comprehensive list of what you need
to know!
- What is the primary purpose of DESeq2 in differential gene
expression analysis?
- To cluster samples based on gene-expression profiles
- To model RNA-seq count data and identify differentially expressed
genes
- To visualize gene-expression data as heatmaps
- To perform gene set enrichment analysis
- In a typical volcano plot, what do the x- and y-axes
represent?
- Log2 fold change and statistical significance (-log10
p-value)
- Mean expression and gene length
- Gene ID and normalized expression
- p-value and sequencing depth
- What type of input data should generally be provided to DESeq2 for
differential expression analysis?
- Raw integer count data
- TPM-normalized expression values
- Log2-transformed expression values
- Z-score standardized expression values
- How does DESeq2 account for differences in sequencing depth and
library composition between samples?
- By applying quantile normalization
- By log2-transforming the counts before statistical testing
- By estimating sample-specific size factors
- By converting raw counts to TPM
- In the context of volcano plots, what does a high absolute value of
log2 fold change indicate?
- Low statistical significance
- High statistical significance
- High gene expression in both groups
- A large difference in gene expression between the compared
groups
- Which of the following is generally the most appropriate statistic
for ranking genes in fgsea?
- Raw p-value
- Adjusted p-value
- Signed t-statistic
- Gene expression level
- When interpreting the results from fgsea, what does a normalized
enrichment score (NES) represent?
- The strength and direction of gene set enrichment
- The proportion of significant genes in a gene set
- The raw p-value for gene set enrichment
- The size of the gene set
- In a DESeq2 results table, what does the
padj column
represent?
- The log2 fold change adjusted for sequencing depth
- The raw p-value adjusted for multiple testing
- The mean normalized expression across samples
- The estimated size factor for each sample
- What is the purpose of including a control group in a DESeq2
analysis?
- To measure the absolute gene expression levels
- To perform hierarchical clustering
- To normalize data for batch effects
- To provide a baseline for comparison of differential expression
- Which fgsea output should generally be used to determine statistical
significance after correcting for testing multiple gene sets?
- Enrichment score (ES)
- Normalized enrichment score (NES)
- Nominal p-value
- FDR-adjusted p-value (padj)
Self-evaluation quiz answers (click to expand)
- b
- a
- a
- c
- d
- c
- a
- b
- d
- d
Curriculum (click to expand)
Introduction
to the DESeq2 R package
DESeq2
tutorial
Introduction to
volcano plots
Introduction to gene
set enrichment analysis (GSEA)
GSEA
tutorial using the fgsea R package
Day 3: Computational precision diagnostics: molecular subtyping of
cancer – part 1
When: Tuesday, 15 September 2026, 13:00–17:00
What: Lecture, Q&A, exercises, and self-study
Program
13.00 - 13.45: Lecture: Molecular Subtyping of Cancer Using Gene
Expression Data
13.45 - 14.00: Break
14.00 - 17.00: Exercise:
Subtyping of cancer- 1; Self study
Today’s learning objectives
After today’s session and going through the reading materials, you
will be able to:
- Understand the basic concept of molecular subtyping of
cancer.
- Describe the concept of distance between vectors.
- Apply Euclidean distance to an expression matrix.
- Understand and apply distance-to-centroid classification on
microarray data.
- Understand and apply the k-nearest neighbor (kNN) classification
algorithm on microarray data.
- Understand and apply single-sample gene set enrichment analysis
(ssGSEA) on microarray data.
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to know
about these topics. It is not a comprehensive list of what you need to
know!
1. What does molecular subtyping of cancer primarily
involve?
a) Identifying every genetic mutation within a tumor
b)
Determining the anatomical stage of a tumor
c) Classifying tumors
into distinct subgroups based on molecular characteristics
d)
Selecting treatment based only on the tissue in which the tumor
originated
2. In K-nearest neighbors (KNN) classification, what
does the value of K represent?
a) The number of features used for
classification
b) The distance metric used to calculate similarity
between data points
c) The number of neighbors considered for
classification
d) The confidence level of the classification
result
3. In distance to centroid classification, how is the
class label of a new data point determined?
a) By comparing the
distances of the data point to all other data points
b) By
calculating the average distance of the data point to all other data
points
c) By assigning the class label of the centroid closest to
the data point
d) By considering the distances of the data point to
all the data points of all classes
4. Which of the following is
a commonly used approach for molecular subtyping of cancer?
a)
Immunohistochemistry (IHC)
b) Next-generation sequencing (NGS)
c) Gene expression profiling
d) All of the above
5. What is
the primary objective of single sample gene set enrichment analysis
(ssGSEA)?
a) To identify differentially expressed genes in multiple
samples
b) To determine the functional enrichment of gene sets in a
single sample
c) To quantify the expression levels of individual
genes in a given sample
d) To measure the degree of correlation
between gene expression patterns in a single sample and predefined gene
sets associated with specific biological functions or pathways
6. What is the CIT subtyping scheme?
a) A machine learning method
for Centralized Intelligent Training
b) A framework defining
Clinical Individual Treatments
c) A data set for Checkpoint
Inhibitor Therapy
d) None of the above
7. Why are we using
leave-one-out for cross-validation instead of the more common five-fold
cross-validation? a) Because the training data set is quite small, and
if we leave out too much of it, we risk changing the very definition of
the subtypes
b) Because it is computationally a lot faster to leave
one sample out, than to leave out a fifth
c) Because using
leave-one-out minimizes the false positive rate
d) All of the
above
8. Why is it still useful to learn how to analyze
microarray data despite the widespread use of RNA-seq?
a)
Microarrays generally provide better gene expression measurements than
RNA-seq
b) Many established cancer subtyping methods and historical
datasets are based on microarray data
c) Microarrays can detect more
transcripts than RNA-seq
d) Microarray data do not require
normalization
Self-evaluation quiz answers (click to expand)
- c
- c
- c
- d
- b
- d
- a
- b
Curriculum (click to expand)
Paper
describing molecular subtyping of cancer
Paper
describing bioinformatics pipelines for molecular subtyping of
cancer
Your turn! Find resources about the following
concepts:
* Euclidean distance
* Distance to centroid
classification
* k-nearest neighbor classification (kNN)
*
Single sample gene set enrichment analysis (ssGSEA)
* The GSVA R
package for performing ssGSEA
* The concept of cross-validation in
machine learning (we will use “leave one out”)
Day 4: Computational precision diagnostics: molecular subtyping of
cancer - part 2
When: Tuesday, 22 September 2026, 13:00–17:00
What: Guest Lecture (Prof. Caroline Ford), Q&A,
exercises, and self-study
Program
13.00 - 14.00: Guest Lecture: Precision Medicine in Women’s Health:
From Research to Clinical Practice
14.00 - 17.00: Exercises:
Subtyping of cancer- 2; Self-study)
Today’s learning objectives
After today you will be able to:
- Distinguish between technical
and biological variations between samples
- Account for technical
variation and relevant biological confounders between samples
-
Estimate tumor purity using the ESTIMATE algorithm
- Critically
assess the impact of these adjustments on cancer subtyping
-
Understand how rank-based similarity measures such as Kendall’s tau can
be used when comparing molecular profiles
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to know
about these topics. It is not a comprehensive list of what you need to
know!
1. What are batch effects in gene expression data?
a)
Random fluctuations in gene expression levels within a sample
b)
Differences in gene expression between different biological
conditions
c) Systematic variations in gene expression attributed to
technical factors
d) Genetic mutations affecting gene expression
patterns
2. When comparing gene expression data of cohorts of
patients, why is it important to consider biological confounders?
a)
Biological confounders can introduce noise in the gene expression
data
b) Biological confounders can lead to biased interpretation of
gene expression differences
c) Biological confounders can affect the
reliability of gene expression measurements
d) All of the above
3. How is the distance between two ranked vectors typically
measured?
a) Hamming distance
b) Euclidean distance
c)
Kendall’s tau distance
d) Mahalanobis distance
4. How does
the ESTIMATE algorithm infer tumor purity from gene expression data?
a) It calculates stromal and immune cell enrichment scores and uses
these to estimate the proportion of tumor cells
b) It directly
identifies and counts tumor cells from gene expression data
c) It
infers copy number variation from microarray expression profiles
d)
It identifies somatic mutations and compares them with known
cancer-driver mutations
5. Should one adjust for tumor purity
before subtyping? Why/why not?
a) Yes, because tumor purity can
differ across breast cancer samples for reasons relating to the
surgery
b) No, because the ESTIMATE algorithm only estimates tumor
purity and is not accurate
c) Yes, because we are only interested in
the signal from the tumor cells as these are the drivers of the
disease
d) No, because the entire tumor microenvironment is what
defines the disease
Self-evaluation quiz answers (click to expand)
- c
- b
- c
- a
- d
Curriculum (click to expand)
Paper
describing how to make RNA-seq backwards compatible with microarray
data
Paper
describing prediction of tumour purity and stromal and immune cell
admixture from expression data
Paper
describing bioinformatics pipelines for molecular subtyping of
cancer - This is same paper as last week, but re-read with a
focus on the n+1 sample integration approach and think about why we did
it differently in this paper than we did in the paper above, where we
used the tool called ComBat. You will need to consult Supplementary
Figure 4 (see under “Additional Information”). Also focus on the effects
of tumor purity. See Supplementary Figure 1
Tumor
microenvironment and tumor heterogeneity (read introduction, review the
video and Figure 1)
Your turn! Find resources about the
following concepts:
* Kendall Tau distance between ranked
vectors
* How to apply the ESTIMATE algorithm using the “immunedeconv”
package
Day 6: Predicting prognosis and treatment response: checkpoint
inhibitor therapy response
When: Tuesday, 06 October 2026, 13:00–17:00
What: Lecture, Q&A, exercises, and self-study
Program
13.00 - 14.00: Lecture: Introduction to checkpoint inhibitor therapy
14.00 - 14.15: Break 14.15 - 17.00: Exercises: checkpoint inhibitor
therapy; Self-study
Today’s learning objectives
After today you will be able to:
- Understand and apply the
following concepts for prediction of response to checkpoint inhibitor
therapy:
* Tumor mutational burden
* T cell Inflamed
gene expression profile
* PD-L1 expression
- Understand
the concepts of checkpoint inhibitor resistance, and why the three
features mentioned above are not enough to fully predict response
-
Understand and apply the tool ESTIMATE for estimating tumor purity
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to know
about these topics. It is not a comprehensive list of what you need to
know!
1. Which of the following is a commonly targeted immune
checkpoint in cancer therapy?
a) PD-1
b) HER2
c) EGFR
d)
BCR-ABL
2. How do immune checkpoint inhibitors work in cancer
therapy?
a) They directly kill cancer cells.
b) They release the
“brakes” of the immune cells and stimulate them to recognize and attack
cancer cells.
c) They prevent the formation of blood vessels that
supply nutrients to tumors.
d) They interfere with the division and
growth of cancer cells.
3. Which of the following is a
well-established mechanism of resistance to immune checkpoint inhibitor
therapy?
a) Loss or impairment of tumor antigen presentation
b)
Increased activation of tumor suppressor genes
c) Increased
infiltration of functional anti-tumor T cells
d) Increased
presentation of tumor neoantigens
4. Which of the following
factors can be used to predict whether a patient will respond to immune
checkpoint inhibitor therapy?
a) Expression levels of immune
checkpoint proteins in tumor cells
b) The inflammatory environment
in the tumor
c) Tumor mutational burden
d) All of the above
5. What is the tumor microenvironment?
a) The size and location
of the tumor within the body
b) The genetic composition of cancer
cells within the tumor
c) The surrounding cellular and non-cellular
components in the tumor’s vicinity
d) The stage of cancer
progression and metastasis
6. What is tumor heterogeneity?
a) The ability of a tumor to metastasize to distant organs
b) The
genetic stability of cancer cells within a tumor
c) The presence of
diverse cell populations with distinct genetic and phenotypic
characteristics within a tumor
d) The responsiveness of a tumor to
chemotherapy or radiation therapy
7. What is tumor mutational
burden (TMB)?
a) The size of the tumor within the body
b) The
number of mutations found in a tumor’s DNA
c) The rate at which a
tumor is growing
d) The presence of specific genetic mutations in
tumor cells
8. Why is TMB relevant in the context of checkpoint
inhibitor therapy?
a) It determines the potential number of mutated
proteins in the tumor that the immune system can recognize as
foreign
b) Highly mutating tumors are more unstable and easier to
kill
c) It is the target of checkpoint inhibitor antibodies
d)
All of the above
Self-evaluation quiz answers (click to expand)
- a
- b
- a
- d
- c
- c
- b
- a
Curriculum (click to expand)
Brief
introduction to immune checkpoint inhibitor therapy
Resistance
to immune checkpoint inhibitor therapy (read introduction and review
Figure 1 + Table 1)
Paper describing
predictions of response to immune checkpoint inhibitor therapy using a T
cell inflammation signature (read introduction + section “2.3.1
T-cell–inflamed GEP”)
Your turn! Find resources for these
concepts:
* Tumor mutational burden (TMB) + how it is measured
*
Besides TMB and the T-cell-inflamed gene expression profile, there is
another clinically used predictive biomarker for response to checkpoint
inhibitor therapy. Please find out what it is.
Day 7: Predicting prognosis and treatment response in checkpoint
inhibitor therapy: Survival Analysis and Biomarker Evaluation
When: Tuesday, 20 October 2026, 13:00–17:00
What: Lecture, Q&A, exercises, and self-study
Program
13.00 - 14.00: Follow-up on day 6 exercises + Q&A; Lecture:
Checkpoint inhibitor therapy resistance
14.00 - 14.15: Break
14.15 - 17.00: Exercise: Analysis of survival with treatment;
Self-study
Today’s learning objectives
After today you will be able to:
- Understand and apply survival
analysis in R, including log-rank test and Kaplan-Meier plots
-
Understand the difference between a prognostic and a predictive
biomarker
- Describe the process of defining predictive biomarkers,
from discovery to validation
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to know
about these topics. It is not a comprehensive list of what you need to
know!
1. What does a Kaplan-Meier plot display?
a) The
average growth rate of cancer cells over time
b) The probability of
survival over time for a group of patients
c) The distribution of
gene mutations in a population
d) The results of a drug trial on a
specific date
2. In a Kaplan-Meier plot, what does a steep drop
in the survival curve indicate?
a) An increase in the probability of
patient survival
b) A higher incidence of adverse events among
patients
c) A significant number of patients surviving longer than
expected
d) A high number of patients experiencing the event being
studied (e.g., death or recurrence) in a short period of time
3. You are analyzing the survival of two groups of cancer patients using
a Kaplan-Meier plot. One group shows a survival curve that consistently
remains higher than the other group’s curve. What conclusion can you
draw from this result?
a) The group with the higher curve has a
significantly lower overall survival rate.
b) The group with the
higher curve has a higher probability of surviving for a longer period
compared to the other group.
c) Both groups have equal survival
chances, but the results are due to random variation.
d) The group
with the higher curve has more patients who withdrew from the study
early.
4. What is the primary purpose of the log-rank test in
survival analysis?
a) To estimate the median survival time of
patients
b) To compare the survival distributions of two or more
groups
c) To determine the accuracy of a survival model
d) To
calculate the hazard ratio between two groups
5. What is the
tumor microenvironment?
a) The size and location of the tumor within
the body
b) The genetic composition of cancer cells within the
tumor
c) The surrounding cellular and non-cellular components in the
tumor’s vicinity
d) The stage of cancer progression and
metastasis
6. In survival analysis, what does censoring
mean?
a) Removing observations that do not fit the statistical
model
b) Excluding patients with extreme survival times
c)
Recording an event that occurs before the end of follow-up
d) The
exact event time is unknown because the event has not occurred by the
end of follow-up or the patient is lost to follow-up
7. What is
a key advantage of the Kaplan–Meier estimator for survival analysis?
a) It estimates survival probabilities while accounting for censored
observations and different follow-up times
b) It estimates the
instantaneous rate at which events occur
c) It assumes that the
event rate is constant over time
d) It estimates the effect of
multiple predictors on survival simultaneously
8. You are
comparing survival times between two groups using the Kaplan-Meier
method. Group A shows a higher survival curve compared to Group B. You
apply a log-rank test and obtain a p-value of 0.06. Based on the
standard 0.05 significance level, what should you conclude?
a) There
is strong evidence that survival times are different between Group A and
Group B.
b) There is weak or no evidence of a statistically
significant difference in survival times between Group A and Group
B.
c) Group B has a significantly higher survival rate than Group
A.
d) Both groups have identical survival distributions.
Self-evaluation quiz answers (click to expand)
- b
- d
- b
- b
- c
- d
- a
- b
Curriculum (click to expand)
Paper about the basic
concepts of survival analysis
How to perform Cox
regression and make Kaplan Meier and forest plots in R How to perform the
log-rank test in R
Your turn! Read up on the following
concepts:
* What is the difference between a predictive and a
prognostic biomarker?
* Get acquainted with the “survival” package
for R
Day 8: Computational precision therapeutics: evaluating targets for
chimeric antigen receptor therapy - part 1
When: Tuesday, 27 October 2026, 13:00–17:00
What: Guest Lecture (Prof. Lars Rønn Olsen), Q&A,
exercises, and self-study
Program
13.00 - 14.00: Guest Lecture: Introduction to chimeric antigen
receptor therapy and online bioinformatics tools for assessing CAR
therapy targets
14.00-14.15: Break
14.15 - 17.00: Exercises:
Assessment of chimeric antigen receptor therapy targets 1;
Self-study
Today’s learning objectives
After today you will be able to:
- Understand the general concept
of chimeric antigen receptor (CAR) therapy
- Understand the features
of a good CAR target, including:
* Target expression on tumor
cells and not on healthy cells
* Target protein subcellular
localization
* Target protein membrane topology
*
Target protein structure
- Understand what DeepLoc2 is used for
- Understand what DeepTMHMM is used for
- Understand what AlphaFold
is used for
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to know
about these topics. It is not a comprehensive list of what you need to
know!
1. How does chimeric antigen receptor (CAR) therapy
work?
a) CARs are synthetic receptors that directly kill cancer
cells by inducing apoptosis.
b) CARs are engineered receptors that
bind to specific antigens on cancer cells and activate the immune system
to attack them.
c) CARs stimulate the production of antibodies that
target cancer cells.
d) CARs directly inhibit the growth and
division of cancer cells.
2. What are B-cell epitopes?
a)
Regions of an antigen that are recognized by antibodies or B-cell
receptors
b) Regions of an antibody that bind to an antigen
c)
Peptide-MHC complexes recognized by T-cell receptors
d)
Transmembrane regions that anchor proteins to the cell membrane
3. Why is CAR therapy targeting CD19 effective despite CD19 being
expressed on normal cells?
a) CAR therapy does kill healthy cells
expressing CD19, primarily B cells, the loss of which can be managed
with immunoglobulin replacement therapy.
b) Normal cells expressing
CD19 have mechanisms to evade CAR-mediated killing
c) CAR therapy is
designed to spare normal cells by modulating the binding affinity to
CD19
d) Normal cells expressing CD19 are replenished through the
body’s natural regenerative processes throughout therapy
4. Why
does CAR therapy resistance occur?
a) Low PD1 expression
b) Low
T cell inflamed gene expression signature
c) Low tumor mutational
burden
d) The tumor cells expressing the target are killed under
selective pressure, leaving only clones that do not express the
target
5. Which sequencing-based data type can be used to
estimate transcript-level expression of a protein-coding CAR target in
tumor samples?
a) Whole-genome sequencing
b) Single-cell
ATAC-seq
c) RNA-seq
d) Protein structure prediction
6.
Why is it necessary for CAR targets to have an ectodomain?
a)
Ectodomains are the only place on a protein that an antibody can
bind
b) Ectodomains are specific to tumor expressed proteins
c)
Ectodomains are located on the outside of cell membranes, visible to CAR
cells
d) All of the above
7. How does the DeepTMHMM tool
work?
a) It uses a rule-based system to match known transmembrane
patterns in protein sequences
b) It applies a machine learning
model, trained on protein sequences, to predict transmembrane regions
and orientation based on amino acid properties
c) It relies on
sequence alignment with transmembrane proteins from similar organisms to
infer structure
d) It predicts protein structure solely by
identifying hydrophobic regions without additional context
Self-evaluation quiz answers (click to expand)
- b
- a
- a
- d
- c
- c
- b
Curriculum (click to expand)
Chimeric
antigen receptor therapy
This figure
outlining what constitutes a good target for chimeric antigen receptor
therapy
Your turn! Figure out what the following bioinformatics
tools can be used for and how they work (in very broad strokes, not the
math):
* DeepLoc 2.0
* deepTMHMM
* AlphaFold
Day 9: Computational precision therapeutics: evaluating targets for
chimeric antigen receptor therapy - part 2
When: Tuesday, 3 November 2026, 13:00–17:00
What: Guest Lecture (Prof. Lars Rønn Olsen), Q&A,
exercises, and self-study
Program
13.00 - 13.30: Follow-up on day 8 exercises, Q&A
13.30 -
17.00: Exercise: Assessment of chimeric antigen receptor therapy targets
2; Self-study
Today’s learning objectives
After today you will be able to:
- Understand the basic concept
of splice variants/protein isoforms
- Understand the general concept
of chimeric antigen receptor (CAR) therapy
- Understand the features
of a good CAR target, including:
* Target isoforms
*
No off target binding
- Understand what a multiple sequence
alignment does
- Understand what the tool FoldSeek does
Self-evaluation quiz (click to expand)
NOTE This is a subset of what you need to
know about these topics. It is not a comprehensive list of what you need
to know!
- What is alternative splicing?
- The duplication of a gene to produce additional copies in the
genome
- The translation of the same mRNA into different proteins
- The processing of a pre-mRNA transcript in different ways to produce
multiple distinct mRNA transcripts from one gene.
- The modification of a protein after translation
- What are protein isoforms?
- Different forms of a protein that can be produced from the same
gene, for example through alternative splicing
- Identical proteins expressed in different tissues
- Proteins encoded by completely different genes but with similar
functions
- Proteins located in different cellular compartments
- Why is it important to consider different protein isoforms when
evaluating a potential CAR therapy target?
- Isoforms are only relevant for DNA sequencing
- Alternative isoforms cannot affect CAR target recognition
- Different isoforms may have different expression patterns or protein
structures
- All isoforms of a gene are always identical
- Why should potential off-target binding be investigated when
selecting a target for CAR therapy?
- To estimate tumor purity
- To identify similar proteins that may also be recognized by the
CAR
- To identify highly variable genes
- To determine the sequencing depth of the RNA-seq data
- Why might we compare the sequence of a potential CAR target with
sequences of other proteins?
- To estimate tumor purity
- To identify proteins with sequence similarity that may need to be
evaluated as potential off-targets
- To normalize RNA-seq data
- To calculate patient survival
- What does FoldSeek allow us to investigate?
- Similarity between protein structures
- RNA-seq library size
- Tumor purity
- Differential gene expression
Self-evaluation quiz answers (click to expand)
- c
- a
- c
- b
- b
- a
Curriculum (click to expand)
Alternative
splicing
Paper about the role of
alternative splicing in therapy-induced target-loss in CAR
therapy
Paper about
the tool FoldSeek
Your turn! Figure out what the following
bioinformatics tools can be used for and how they work (in very broad
strokes, not the math):
* Multiple sequence alignment and the R
package “msa”
Day 10: Multiple choice test + introduction to project work
When: Tuesday, 10 November 2026, 13:00–17:00
What: Multiple choice test, with no aids - PLEASE BE ON
TIME!
Day 11-12: Project work
When: Tuesday, 17 and 24 November 2026,
13:00–17:00
What: working on your projects
Day 13: Oral presentations
When: Tuesday, 1 December 2026, 13:00–17:00
What: Presenting your project works
