You can either dichotomise the protein expression and tumour grades into low and high groups and use either fisher's exact or chi square test to get a p value. Or, if the data is continuous then plot box and whisker chart against all the tumour grades and rum Mann whitney U test to look for significance. You can also use spearman correlation coefficient if both the variables are continuous e.g.s looking for correlation between expression of one protein vs. the other protein.

For survival analysis you would need to plot Kaplan Meier curves and then use log rank test to get the p values for difference between the two groups, and then Cox hazard model to get the hazard ratio.

You can also do multivariate analysis using cox model.