Numerous disease-causing and neutral variations have been identified in several genes and proteins. To gain insight into the total number of harmful variations we have performed protein-wide predictions for Bruton tyrosine kinase (BTK) involved in primary immunodeficiency and mismatch repair (MMR) system proteins in gastric cancers. Mutation rates have been measured and estimated for certain proteins, however, are not very useful in this respect as the rates are known to vary by 100 fold depending on the gene. We developed dedicated predictors both for the BTK, called PONBTK (1), and for four MMR proteins, called PON-MMR2 (2). The methods were carefully evaluated before used for the analyses. They both have very good performance according to several measures. In the case of BTK kinase domain we found that altogether two thirds of all the possible single-nucleotide substitution-caused amino acid variations (SNAVs) are harmful. In the case of the MMR proteins, where all the possible substitutions i.e. 19 per position, were investigated the ratios are somewhat smaller than for BTK and different for individual proteins. Predictions with PON-P2 (3), a generic pathogenicity predictor, indicate very different ratios of harmful variants in proteins. PON-P2 is very suitable for this kind of analysis due to its high performance and ability to distinguish harmful variants. We have applied this approach to all the 22 mitochondrial tRNA molecules by using our novel PON-mt-tRNA predictor (4). Several mechanisms are behind the harmful variants. When developing a new predictor for protein solubility affecting variants (PON-Sol, 4), we tested the method on interleukin 1-β and found structure context dependent distribution differences for the solubility increasing and decreasing variants. Knowledge about harmful variants in genes and proteins allows deeper insight into diseases and when combined with other information, such as protein structures, understanding the mechanisms behind diseases due to variants. All the tools mentioned are freely available at http://structure.bmc.lu.se/ References 1. Väliaho, J., Faisal, I., Ortutay, C., Smith, C. I. E. and Vihinen, M. (2015) Characterization of all possible single-nucleotide change caused amino acid substitutions in the kinase domain of Bruton tyrosine kinase. Hum. Mutat. 36, 638-647. 2. Niroula, A. and Vihinen, M. (2015) Classification of amino acid substitutions in mismatch repair proteins using PON-MMR2. 3. Niroula, A., Urolagin, S. and Vihinen, M. (2015) PON-P2: Prediction method for fast and reliable identification of harmful variants. PLoS ONE 10(2):e0117380. 4. Niroula, A. and Vihinen, M. (2016) PON-mt-tRNA: a multifactorial probability-based method for classification of mitochondrial tRNA variations. Nucleic Acids Res. 44, 2020-2027. 5. Yang, Y., Niroula, A., Shen, B., and Vihinen, M. PON-Sol: prediction of effects of variants on protein solubility. Bioinf. (in press).

1. Ratio of harmful amino acid
substitutions
Mauno Vihinen
Protein Structure and Bioinformatics Group
Department of Experimental Medical Science
Lund University, Sweden

2. Focus on variation
Data collection
Method performance
assessment
Systematics
Prediction method
development
Predictors
Mechanisms

3. How many variations are disease
related?
Current databases don’t provide the answer
Depends on gene/protein/domain/region
Bruton tyrosine kinase (BTK) in X-linked agammaglobulinemia

4. Amino acid substitutions
Hydrophilic
Hydrophobic Acidic Basic Polar Special
-> A F I L M V W Y D E H K R N Q S T C G P X total
A 0 1 0.6 0.4 0 0 0 0.8 0 2.8
F 0 0 0.4 0.1 0.3 1 0.1 0.1 2.1
I 0.1 0 0 0.1 0 0 0 1 0 0.1 0.6 0 2
L 1.3 0.3 0 0 0.3 0.1 0 0.6 0.1 0.6 3.8 0.6 7.6
M 1.1 0.1 0 0.4 0.4 0.1 2 0 4.2
V 0.4 0.7 0 0 0 0.1 0.4 0.1 0.3 0 0.1 2.2
W 0.1 0.8 0.3 0.3 0 4.1 5.6
Y 0 0 0.7 0.6 0.6 0.8 1.3 4.8 8.7
D 0 0.4 0.1 0 0.1 0.1 0.3 0.3 0 1.4
E 0 0 0.7 0 0.3 0 0.4 1.8 3.2
H 0 0.1 0.1 0 0.4 0 0.1 0.3 0 1.1
K 0.1 0 1.1 0 0.4 0.3 0 0 1.4 3.4
R 0 0.3 0 6.3 4.1 0.6 0.1 4.5 1 0.3 2.5 2 1.1 8.3 31
N 0 0.1 0 0 0.1 0 0 0 0.1 0.4
Q 0 0 0 0.3 0 0.1 0 0.1 6.2 6.7
S 0 0.7 0.1 0.1 0 0.4 0.1 0 0 0 0 0 0.7 1 3.2
T 0.1 0.4 0 0 0 0 0 0 1.1 0 1.7
C 0.7 0.3 1.7 0.4 0.3 0 0.4 0.7 4.5
G 0.1 0.1 0.1 1.1 2.1 1.5 0 0 0.1 0.3 5.6
P 0.3 0.6 0 0 0.3 0 0.7 0.7 0 0 2.5
total 1 3.5 2.1 1.7 0.1 2.5 6.9 2.8 3.6 3.9 5 1.4 5 2.1 4.8 4.8 3.5 4.2 3.5 8 29.5 100

5. Approach
Predictions
Methods of highest possible performance
Machine learning methods
trained with experimentally verified cases, careful selection, representativeness
Benchmarking

6. PON-BTK
More than 500 kinases, BTK contains the largest number of disease-
causing variants among them
Dedicated predictor for kinase domain of Bruton
tyrosine kinase
SNAVs, single-nucleotide substitution caused
amino acid variations
Analysis of all amino acid substitutions caused by
single nucleotide change
67% of variations harmful
Väliaho et al. Hum. Mutat. 2015

7. Experimental studies

8. PON-MMR
Dedicated predictor for a protein complex, mismatch repair (MMR) system
Lynch syndrome and other gastrointestinal cancers
Collection of data from literature, InSiGHT database
785 amino acid substitutions in InSiGHT database in 5 proteins
Clear functional information for disease relevance only for 168
INSiGHT variation interpretation committee rapport (Thompson et al. Nat. Genet 2014)
Classified 1370 variants out of 2360 investigated
46 of those we had predicted to be either benign or pathogenic
44/46 were correct (96%)
Ali et al. Hum. Mutat. 2012

9. PON-MMR, structural verification
Structure of MSH2-MSH6 dimer
Structural explanations match for 105 out of 109 variants

10. PON-MMR2
Novel predictor
for MLH1, MSH2, MSH6, PMS2
Trained with InSiGHT and PON-MMR data
Extensive feature selection and method
training
5 useful features from altogether 624
Accuracy and MCC: 0.84 and 0.70, cross-
validation 0.85 and 0.67 for an independent
test dataset
Niroula and Vihinen, Hum. Mutat. 2015

11. Ratios of harmful variants

12. Mitochondrial tRNA
Human cells have two sets of tRNAs - nuclear and mitochondrial
Almost all pathogenic variations in tRNAs are found in mitochondrial
tRNAs
Evidence-based methods have been developed to classify mitochondrial
tRNA variations
Conservation, Biochemical test, histochemical test, heteroplasmy,
segregation, single fibre, trans-mitochondrial cybrid, etc
About 200 variants were classified by Yarham et al.
Niroula and Vihinen, NAR 2016

13. PON-mt-tRNA
Classification of all possible single nucleotide
substitutions in all 22 human mt-tRNAs
Pathogenic variations are concentrated in the stems
Anticodon loop has the highest frequency of pathogenic
variations among loops
51.0% of all variants predicted to be harmful
42.0% of all variants in yeast arginine tRNACCU (Li et al.
Science 2016)

14. PON-P2
PON-P2 classifies amino acid substitutions into three classes
benign
pathogenic
unclassified variants (UVs)
Machine learning -based method (random forest)
Features
Physical and biochemical properties of amino acids
Gene Ontology annotations
Evolutionary features
Functional annotation
Niroula et al PLoS ONE 2015

15. Variants in cancer
Somatic variations in 7,042 cancer genomes/exomes
Alexandrov et al. (Nature 2013)
30 cancer types
~5 million variants
2.63 million in coding region
824,001 amino acid substitutions
Predicted the impact of AASs using PON-P2
14.24% of all variations were predicted harmful
14.71% of COSMIC variation (total 647,872)
39.88% variations in Cancer Gene Census (CGC) genes
Niroula and Vihinen, BMC Med Genet 2015

16. Summary
Estimates for the ratio of harmful variants
- BTK kinase domain, SNAVs
- MMR system proteins, all variants
- Mitochondrial tRNAs
- Cancer
- 30 cancer types
- COSMIC
- Cancer Gene Census proteins
Highly variable ratios between proteins and domains, and within proteins

17. Thanks!
Abhishek Gabriel Gerard Jelena Siddhaling Yang
Niroula Teku Schaafsma Calyseva Urolagin Yang
Heidi Preethy Jouni
Ali Nair Väliaho
http://structure.bmc.lu.se
mauno.vihinen@med.lu.se