SLIDE 1
Crystal balls in the hospital Predicting treatment response
- Dr. Cecilia Engel Thomas
Post doc in Schwenk lab at SciLifeLab and KTH
Crystal balls in the hospital Predicting treatment response Dr. - - PowerPoint PPT Presentation
Crystal balls in the hospital Predicting treatment response Dr. - - PowerPoint PPT Presentation
Crystal balls in the hospital Predicting treatment response Dr. Cecilia Engel Thomas Post doc in Schwenk lab at SciLifeLab and KTH www.medicalnewstoday.com/articles/315901.php ? Bachelors in Technical Biomedicine Masters in Bioinformatics and
SLIDE 2
SLIDE 3
www.medicalnewstoday.com/articles/315901.php
SLIDE 4
SLIDE 5
?
SLIDE 6
Bachelors in Technical Biomedicine Masters in Bioinformatics and Systems Biology
SLIDE 7
Cancer prediction Text mining of Electronic Patient Records Text mining of scientific literature Metagenomics habitat prediction Clustering of single cell data
SLIDE 8
SLIDE 9
SLIDE 10
Crystal balls in the hospital Predicting treatment response
SLIDE 11
Adapted from Schork 2015
SLIDE 12
Source: https://www.drugs.com/sfx/rosuvastatin-side-effects.html
SLIDE 13
SLIDE 14
Non-responder Responder
SLIDE 15
How to make a “crystal ball” Step 1: Collect information/data on the patients Step 2: Come up with a good way to combine that data and make predictions
SLIDE 16
Step 1: Collect information/data on the patients
SLIDE 17
Genome Epigenome Transcriptome Proteome Metabolome Phenome
Cancer Diabetes Height Psyciatric disease Obesity
Molecular readouts
SNP CNV LOH Genomic rearrangement Rare variant DNA methylation Histone modifjcation Chromatin accessibility TF binding Gene expression Alternative splicing Long non-coding RNA Small RNA miRNA Protein expression Post-translational modifjcation Cytokine array Metabolite profjling in serum, plasma, urine, CSF, etc.
TF miRNA Adapted from Ritchie et al., 2015
SLIDE 18
Step 2: Come up with a good way to combine that data and make predictions
SLIDE 19
Machine learning
SLIDE 20
Machine learning
Input Output/answer
Image sources: http://phdp.github.io/posts/2013-07-05-dtl.html, https://pixabay.com/en/computers-keys-rays-1420200/
SLIDE 21
Input: Movies Mark has watched Output: Movies Mark will like
SLIDE 22
Input: Omics data Output: Response to therapy Who will respond to therapy?
SLIDE 23
Which patients with type 2 diabetes should get which treatment?
SLIDE 24
Prevalence 415 million in 2013 642 million by 2040
Type 2 diabetes
Leading cause of non- traumatic amputations Complex disease with unknown cause Defined by dysregulation
- f sugar metabolism
Higher risk with
- verweight, smoking,
sedentary lifestyle, unhealthy eating habits, and genetics Can lead to blindness, dementia, need for dialysis…
SLIDE 25
Nguyen & Varela 2016
Lifestyle change Medication Bariatric surgery
Image sources: https://www.everydayhealth.com/type-2-diabetes/guide/treatment/
SLIDE 26
For whom can bariatric surgery be a treatment for type 2 diabetes?
SLIDE 27
Image sources: http://www.hormone.org/questions-and-answers/2012/bariatric-surger, http://www.bariatriccarecenter.com/maintain-weight-loss-after-Bariatric-Surgery-california.htm
Nguyen & Varela 2016
Who will respond to therapy?
SLIDE 28
Image sources: http://soonerorlighter.bangordailynews.com. Nguyen & Varela 2016
Diabetes Yes/No
SLIDE 29
Still diabetic/non-responder Not diabetic/responder Prediction algorithm Input: Information about the patient Output: Response to therapy Who will respond to therapy?
SLIDE 30
Still diabetic/non-responder Prediction algorithm
SLIDE 31
Still diabetic/non-responder Prediction algorithm
SLIDE 32
Not diabetic/responder Prediction algorithm
SLIDE 33