1. Does the proposed algorithm make sense to you? Can you follow the logic?

The logic seems okay. I think the amount of pseudocode here is very rich in terms of length. In terms of what we're supposed to be checking for, I think you have a solid handle. I like that things are very stepwise and are laid out very plainly. I did not see you planning to use a dictionary anywhere. I think using one might be helpful for storing information regarding your records. You might be able to use this dictionary to figure out if you have duplicates in "dictionary-checking" loop... That's what I've kind of heard other people chatting about as we try to better define approaches to the deduping problem. I think you have a nice start here and I anticipate this being a good starting point for you to craft your nested set of conditionals.

2. Does the algorithm do everything it's supposed to do?

From what I can tell, yes. Referencing my above comment (1.) provides further detail on why I think this proposed algorithm bears the skeletal structure of what we're going to need to actually code over the next two weeks.

3. Are proposed functions reasonable? Are they "standalone" pieces of code?

Generally, yes, you seem to have crafted some reasonable functions for use. It appears you plan to check for everything we've been instructed to check for-- good. I am a little confused regarding your CIGAR string function though. Make sure that for 'I' you are subtracting and for 'D' you are adding to the CIGAR string (maybe utilize a counter here too in your actual code)... I have some brainstorming here to do myself on how exactly I'm going to write out my actual algorithm. Especially when considering directionality, I believe where we add or subtract from becomes a concern given the plus or minus strand. Likewise, your functions (after being coded), will definitely have the capacity to run as standalone pieces of code.

Nice work, pal! I'm under the impression you're thinking the right thoughts here!

Hi David,

I think you defined the problem well, and have good examples of what a properly de-duped file may look like. I appreciate the level of detail you include in your functions. Something that I worked on this time was including more code snippets and pieces of code I at least thought should work. I'm hoping this will kind of jump start my thinking process when I actually sit down the code and it also helps me think about the problem in a more algorithmic manner when I'm writing my sudocode. It would also be good practice just thinking about python and getting to know the different data structures and variables you will need to solve your problem.

In addition I think should review some of your logic late in your main look. I'm not sure this problem can be solved with sets of conditionals that filter records that way. I would look into dictionaries/sets or some way to reference records you know are unique, so you can ask whether the current record is a member of that list and decide whether to write it or not. But overall nice job.

• Does the proposed algorithm make sense to you? Can you follow the logic?

This algorithm is a little confusing. Here are some adjustments that I think might help.

Use Grep to isolate SAM alignment column 2, the bitwise flag.
Using Grep, acquire the number in column 3 of the alignment section of the SAM
It might be hard and inefficient to go between bash(grep) and python. Maybe you should only stick to python.

If the flag + 16 = 16, the read is a minus read.
You have to use the bitwise(not +) and to check the flags.

For read copied into out file, do the following.
If you copied into out file, how do you skip the duplicates?

Lastly, I think you didn’t have to make that many variables (but not important).

• Does the algorithm do everything it's supposed to do? (see part 1)

I think this algorithm is flawed because it only checks if a read matches the next couple consecutive reads or not. Okay suppose the first read matches the 2nd and 4th read. Then your algorithm will recognize the 2nd one is a duplicate but not the 4th one since the 3rd read and the 4th read don’t match. I recommend doing this with a dictionary with key: things you want to check(strand, chromosome, starting position, UMI)

• Are proposed functions reasonable? Are they "standalone" pieces of code?

The Plus_adjust_start and Minus_adjust_start functions are very clear!

Maybe you could add some code to check UMI against UMI_List in your Umi_Checker function.

In all, this is a good start! You demonstrate good understanding of what a duplicate is and how to detect, however, the logic doesn’t check out.