Category: Raw DNA Phasing

In my last Blog, I said that I would work on the Maternal Patterns and then fill in blanks. However, my Maternal Pattern Table is not very filled. After some thought and re-reading the Whit Athey Paper on Phasing, on which I base this work, I decided to:

1. Fill in the Paternal Blanks
2. Use the Paternal Data to fill in the Maternal alleles
3. Fill out the Maternal Pattern Table
4. Fill in the Maternal blanks based on the Maternal Pattern Table

Filling In Paternal Blanks

I might as well start filling in the AAAAA Patterns. On my Dad Pattern Excel Spreadsheet, I can filter for that pattern:

However, I now need a formula for Excel including all the ID positions above. This was the point of my starting this project over – to get those IDs. The formula will be in the form of “Between A And B OR Between C And D OR…” So first I need a formula in Excel to create the formula in Access. That formula is called Concatenate. According to a Google search, concatenate means to “link (things) together in a chain or series”. The symbol in Excel for concatenate is simply the ampersand (&).

Here is my formula and the outcome:

However, I have another idea. I can concatenate the concatenation. First, I add an extra space on the end of my “Or”. Then I drag down the formula to fill in the other chromosomes. Then I take off the last “Or”.

That gives me this helpful string of AAAAA Positions:

This will save me a lot of cutting and pasting in Access.

Back to Access

First I copied my old Table to a new one called tblFillInOne. I will create an Update Query for that Table.

I am only updating Dad alleles from other Dad alleles, so I import those 5 alleles plus the location ID. Then I use the expression builder, to paste in the location of the AAAAA Patterns in all 22 chromosomes. So now I have the Pattern and the location, but I need some more criteria. I would like the criteria to say if there is any allele in any of the five columns and any blanks in those columns, then replace the blank space with one of the existing alleles.

Here is a simple Update Query:

This says, that if my allele is null and Sharon’s isn’t, then replace mine with Sharon’s. The problem is that this would take four separate Update Queries. With 5 siblings, that would be 20 queries.

Another risky Update Query would use this form:

Here I am saying if any allele is not null (other than mine) replace that in my slot where I have a blank. The thing I don’t know if the Update To: field can have a variable criteria. I’ll try it. When I run this as a Select Query, it puts a bit of a strain on my computer. Eventually, it gives me 18,385 rows. When I run the View function on the Update Query, I get the same number of rows, so I’ll hit the Run button and hope for the best.

If I run this Select Query again, I should get no results if everything updated.  I did get no results, so I assume that it worked. I want to save this Update Query and use it for the other four siblings.

Filling in Sharon’s missing alleles from the AAAAA Paternal pattern

I used the same logic for Sharon:

Now she has all the Is Null values and I don’t. I moved the Update To: criteria over to Sharon. I took out Sharon’s allele and added mine in her place. Again, this gives my old computer a workout. I get 18,315 rows again which seems suspicious. I see the problem. I appears that Access updated my results with a (-1) rather than with an allele.

That means that I just have to do 20 Queries. However, they should go quickly.

Back to the Simple AAAAA Query

Due to all the Update Queries, I’ll make a Spreadsheet to keep track of each Update Query I do:

It turns out that it is easier to run this Update Query sorted by ‘From’:

That way, I can just move Sharon’s allele from Dad and the Is Null along the Update Query:

With these fast 20 Update Queries, I updated over 100,000 alleles:

AAAAB Fill-in

This could be a little easier. For this one, we don’t want to touch the last ‘B’. The last B represents Lori, so we will only be filling in to and with the other four siblings.

And then we need the fill-in locations.

AAABA and AAABB Fill-in

AAABA is about the same as AAAAB except the B in the AAABA corresponds to Jonathan. He is all alone as a B so he gives no alleles and takes no alleles. The other siblings share their AAAA’s in this Pattern.

In an AAABB Pattern, the three A’s will share with each other and the two B’s will share with each other. This happens to break down along V1 and V2 lines, so I expect there will not be as much sharing as between AncestryDNA versions.  The sharing of A’s and B’s looks like this in my Fill-in Tracker:

I have darkened out the areas where an A cannot share their A with a B and a B cannot share their B with an A. As I predicted, the AAABB filled-in alleles were less:

All the other patterns filled in

All the other patterns will be of the same type. There is one AAAAA which is all the same. The other combinations are four of one type and one of the other or three of one type and two of the other.

There are 20 fill-in’s for AAAAA. As a quality check, there are 12 fill-in’s for a 4-1 Pattern and there are eight fill-in’s for a 3-2 Pattern. I would recommend using a fill-in tracker to make sure all the combinations are being covered. The specific numbers of alleles being filled in for each combination of each Pattern are not all that important, but they are interesting.

Fixing an abbab mistake

When I was filling in the ABBAB Pattern, I noticed a mistake I made. I filled in 3754 rows of Joel alleles into Heidi blank spots. In an ABBAB Pattern, I am only supposed to be filling in my alleles into Jon’s blanks. Here is the mistake:

That means in those positions, I’ll have an ABAAB Pattern where I should have an ABBAB Pattern. Oh no. So how do I fix that? I need a fix query. Under Pos ID, I’ll put in all the locations that are supposed to be ABBAB. Then I’ll make sure the first position isn’t the same as the second:

That results in only 103 rows.

If I update those 103 rows to Null, that should be a start:

Next I set the first position to be different than the last in this ABBAB Pattern:

That fixes another 212 rows. That may be all the rows to fix. I looked for more JoelfromDad = Heidi from dad where JoelfromDad <> LorifromDad and where JonfromDad <> LorifromDad, but didn’t see anything. The other updates must have been in areas with AAAAA by chance areas. In the meantime, I copied the first two columns on the left to the right, so I don’t lose my place when I am scanning across the spreadsheet.

Dad Pattern Fill-in First Round

The dark blue areas are the ones where there should not be any filling in based on the Pattern.

Summary

• The Fill-in Step is a major part of phasing. In this step I filled in over 1 million paternal alleles in my DNA and in my 4 siblings’ DNA.
• I noticed a mistake I made along the way, but figured out a way to fix it.
• I figured out a shortcut to describe the different patterns by way of ID’s. The shortcut involved using a concatenation of a concatenation.
• I haven’t yet filled in the random AAAAA Patterns that are within the other patterns. I imagine that would be important to do at some point. I know that David Pike has a utility to find Runs of Homozygosity. I suppose that would be useful for filling in alleles.

In my last Blog, Part 4, I found that I needed to go back to improve a method from an earlier step to make a later step work much easier. This did two things:

1. Gave me a cleaner database
2. Set me back a ways

Re-do Principle 1: Homozygous Siblings

I need now to create a new table. This will have the allele from Mom and Dad for each sibling. I copied my previous table to a new one called tblV1andV2HomozygousSibs. I opened my new table in design view and added the 10 new fields that I needed:

The first five of the new fields will be have the Mom Patterns and the last five will have the Dad Patterns. Right now they are just blank. I’ll use an update query to add in homozygous alleles:

This query says when my allele 1 is the same as allele 2 (homozygous), put allele 1 into the slot from my Mom and my Dad. The Dad slot goes off the page, but is there. When I run the update, it fills in over 485,000 lines. To do this by hand would have taken a while. This is the first step to filling in the Mom and Dad Patterns:

I do the same query for each of my four other siblings. Care needs to  be made the the right alleles are going in the right place. For example I wouldn’t want to put a Lori allele into a JonfromDad column. Then I check to see if the columns are filling in:

If I recall right, this step fills in (or phases) about 8 million alleles. We don’t see any patterns yet other than AAAAA, but patterns are emerging in other parts of the table.

Step 2 – Homozygous Mom

Here when mom has a GG for example, she would have to give a G to each child at that position, as that is all she has. I’ll use the Update Query again for this. Here is the Criteria:

Here is the Update part:

Step 3 – Heterozygous Siblings

Here is an example:

I have TC in my two alleles at this position as do my siblings. My mom must have had CC as she gave a C to each of her siblings. That leaves a T that we must have gotten from our dad. It looks like I may need 10 Update Queries for this one. Here is the criteria:

The query says that Joelallele1 is not the same as Joelallele2 (Heterozygous Sibling) and I received my allele1 from mom.

I update the table to say I got my other allele from my Dad. This is a little more complicated Update Query. I then reverse the Joelallele 1 and 2. When I get allele2 from mom, I get allele1 from Dad. Before I run the Update Query, I view it each time to see if there is a reasonable number of rows being updated. If no rows are updated, there is probably an error in my query. This update is in the 40-50,000 row range. Also, if I get values in the view panes, it often means I have put the results in the wrong field. Usually many empty rows in the view output is a good thing.

I forgot to copy and rename my Homozygous Sibs Table, so I just renamed it to tblV1andV2w3Principles.

Finding Patterns

This time, I want to add ID’s to my patterns, so I’ll add two columns to my old Pattern Spreadsheet in Excel:

Rather than do formulae for each pattern again, I’ll just scroll through my table to see if I can finesse the Pattern boundaries and add Position IDs.

Finessing Pattern Boundaries

Here is an example at Chromosome 1 in the 77M range. There I had a change from ABBBB to ABABB. In my previous query, I only looked at Dad patterns where all the alleles were filled in. However, in the original pattern, we can infer the pattern even when alleles are missing.

Previously, I had the change at the top row where there is a full pattern. However, in going from ABBBB to ABABB, we only need the first three positions to identify the pattern. And actually, we only need position 2 and 3 to identify ABABB. At ID 25839, there is an AGG??? Pattern. This has to be in the form of ABBBB. Then 4 lines later, is ?AG??. This has to be an ABABB Pattern. Here is how I noted the change in the 77M range on my Excel Dad Pattern Spreadsheet:

The DadStart and DadEnd columns have the refined Position numbers.

Refined Chromosome 1

• I had noted previously a possible AAAAA Pattern between AAAAB and ABBBB. It turns out that that is required. This is because to go from AAAAB (same as BBBBA) to ABBBB would require two changes. Only one change is allowed at a time. I will need to fill in the Positions and IDs.
• The three ABABA Pattern areas need to be combined into one. They occur in a Centromere and in an excess IBD area. The Genealogy Junkie has a good Blog on that topic. I downloaded a file she had with the exact areas.
• I added the IDs for the start and stop of the Chromosome as tested as well as the start of the next Chromosome. These are highlighted in dark purple.
• Only 22 chromosomes to go.

Chromosome 2 Refined

Here I added a new column. This is the number of IDs or SNP positions between patterns. Note that there is a negative 4 in one case. This was an odd case where the two patterns at the crossover were inverted. I didn’t know what to do there, so I left it as is. There is a Centromere from 92-95M, so I will combine the two AAAAB Patterns that I have when I create the clean version of this table.

Chromosome 4 refined

Here I had to add a green AAAAA Pattern to make this work. Note that I am getting fewer crossovers.

Chromosome 5

Here is another case where an AAAAA Pattern is needed:

The pattern is needed between AABAA and AAAAB for two reasons. For one, there is a large gap between the end of AABAA and the beginning of AAAAB. Also, to go from AABAA and AAAAB requires two changes and only one is allowed. That requires an intermediary step of AAAAA between these two patterns [AABAA > AAAAA > AAAAB].

Here is Chromosome 5 completed:

The addition of the AAAAA Pattern results in the addition of two crossovers. Another note is that I could have had the first pattern start at the beginning of the Chromosome and have the last pattern end at the end of the chromosome. That is because there is not much room there for other crossovers.

a chromosome 8 Decision

The issue here is the two AAAAB Patterns in a row. Should they be combined or should I add an AAAAA Pattern between the two AAAAB’s? I’m going with combining. The reason is that if I put an AAAAA between the two, that would give Lori two paternal crossovers in a fairly short span. This does not happen in nature – at least in the middle of a chromosome. This would be like inheriting a 2 cM segment from a grandparent.

Chromosome 9 decision

Lori has two crossovers in a row, which is not ideal. Then there are two ABAAA patterns in a row. I decided to combine these. This is because when I look at the table, there is a centromere in there and a lot of missing SNPs. If I did create an AAAAA pattern, that would result in two close crossover for Sharon.

Here is the cleaned up version with the rogue SNPs taken out:

Missing Pattern Chromosome 10

There is a missing pattern between Lori and Jonathan’s first crossovers. AABAA > ABAAA is two changes, so I need to insert an AAAAA Pattern between the two. This will result in two new crossovers: one for Heidi and one for Sharon.

Chromosome 11 – Halfway?

The good news is that I’m at about 2/3 of the way. I have over 900,000 locations and Chromosome 11 brings us past the 600,000 mark. Note again the need for an AAAAA Pattern between the last two patterns. That will add a Lori crossover and a Jonthathan crossover, so they won’t be left out.

Chromosome 12 patterns

Chromosome 12 looks like it is missing a lot betwee ABBBB and AABAA. However, it is just missing an AAAAA. That is because ABBBB is the same as BAAAA. The progress goes BAAAA > AAAAA > AABAA. As it turns out the crossovers that have to do with transposing relate to me (Joel). The extra crossovers go to me and Heidi.

Here are the numbers filled in for Chromosome 12:

Sketchy Chromosome 13

I note that Chromosome 13 is a bit sketchy, with no identified sibling crossovers. It appears that AAAAA Patterns are needed here also.  There is about a 4M space where there is room for an AAAAA Pattern between 24M and 28M. There is also room after the AAAAB Pattern which would give Lori another Paternal crossover. This last crossover is shown in Gedmatch:

These are matches of my father’s first cousin to myself and four other siblings. This shows Lori’s crossover on the bottom match. As all siblings match to the end of the Chromosome, that would be the AAAAA Pattern.

Here is the finished Chromosome 13:

Lori’s crossover as shown in Gedmatch shows on my table at about 90M. Keep in mind that Gedmatch uses Build 36 and my table is in Build 37.

Chromosome 21: Refinement Example

Here is an example of a refinement. In my initial query, I was looking for patterns that were filled. However, in going from AABBA to ABAAB (which defines my crossover), it is the same as going from BBAAB to ABAAB. The only change in pattern is in the first three letters: BBA to ABA. We can see that change here even though the last three letters are missing:

Chromosome 22: Extra AAAAA needed

On Chromosome 22, there is a lot of room at the beginning of the Chromosome to put in an AAAAA Pattern:

There are about 5M SNPs between the start of the Chromosome (16M) to where the AAABA Pattern starts at 21M. I have 4 of my siblings mapped out using visual phasing:

This shows on the paternal side (Frazer) that there is an AAAAA Pattern. That is represented above at the start of the Chromosome in blue. I am just missing Lori. Without looking at all her results, I see she has a full match with Heidi at the beginning of the Chromosome:

And here is the last Paternal Chromosome finished:

It was a lot of work, but now I have what should be the start and stop points for all the Paternal segments for me and my four siblings.

Summary

• I needed sequential IDs for my Access Queries to fill in missing alleles
• To do this I needed to go back to the beginning and re-import the raw data for six people
• I created a table for five siblings showing where they got their paternal and maternal alleles based on three principals.
• I went back to my Paternal Pattern Table and refined what I had already done
• I also added IDs to my Paternal Pattern Table
• Next up is to look at the Maternal Pattern Table and start filling in blanks using MS Access

In my last Blog, I looked at Maternal and Paternal Patterns created by initial phasing of my raw DNA. The Paternal side patterns were pretty complete, but I didn’t have much on the Maternal side. I summarized the patterns I found in a spreadsheet. I added siblings’ crossovers where found and added start and end positions on each chromosome where found.

Filling in the Blanks

• First I made copies of my Mom and Dad Spreadsheets
• Then I cleaned them up, taking out the Patterns that were only at one location (one SNP)
• Then I used a filter to get the patterns.

The AAAAB Pattern filter on the Paternal Patterns spreadsheet looks like this:

On the Maternal Table, it looks like this:

I’ll start with the maternal Patterns as it will be easier. First I copied that Table I was working on to create a new Fillin Table called tblFillinStep1.

Here is what I portion of the Chromosome 10 AAAAB Maternal Pattern looks like:

In my previous analyses before I had 5 siblings tested, I had sequential ID’s. This made things easier in choosing patterns. Now, however, I am having trouble getting the ID’s sequential. This is probably due to my adding the extra alleles. My options now are to update the missing alleles by hand or create queries to update them.

Just Like Starting Over – a New Access Database

I decided to try importing all the raw DNA files into a new Access Database. This time I won’t let Access assign the key data field. I’ll use the rsid as the key data field. Perhaps then I can assign a new ID that will apply to the merged V1/V2 AncestryDNA dataset.

One problem that I’ve noted with the AncestryDNA downloads is that they only give you a date on the filesname. There is no indication in the file or in the file name of who the data is for. That means that it is important to rename your files. I uploaded my sister Heidi’s results to Gedmatch on 26 April 2015, so that is a clue. I see an AncestryDNA zip file from 25 April 2015, so that is a clue. I guess I’ll add her name to the file now!

The Clean start

Here is the Access Database with just 5 tables:

Here’s my Query to create an AncestryDNA V1 Raw Data Table:

I see in my results, I got more lines than last time. I checked and I forgot to only include Chromosome 1-23. I corrected that and got the 700,153 lines as before. I did and analogous query for Jon and Lori and got 666,532 rows.

All V1 Results plus the v2 results where they are they match v1

Next, I’ll create a Query between the two tables I just made.

This will have a right hand join. As stated above, it will produce all the V1 data and only those records from V2 where they equal V1. That brings in Jonathan and Lori into the V1 results. I made this data into a new table.

Finding v2 data not in v1

Next I want to add only the V2 part that isn’t in V1. When I get this, I can add this to my query to get all the results. This is the query that I couldn’t remember how to do last time. I put V2 on the left, with a right hand query.

This gives everything from V2 plus the V2 that equals V1. However, the trick is that I set V1 to ‘Is Null’ which takes out the V1. That should give only the right section of the peach colored circle below.

For some reason, my new number for the right entire circle is 666,532. The new query results is the right hand circle minus the overlapping data which is 242,494. I’ll append this query to the table I made with my previous Query. I renamed this table as tblV1andV2 and it has 942, 647 rows.

Next, I want to sort the table and add an ID.  I copied the structure of the table to a new table. Then I added a Position ID (Pos ID) with an autonumber. Then I made a query to append the old data to the new table with the autonumbered Pos ID. This gave me what I wanted but not in the right order. So then I used the sort function to get the table in the shape I wanted:

I can tell the table is right by going to the last record:

Before the last record would include the V2 alleles that were added but they weren’t sorted. This long process gives me a sorted ID in the same relative position as the sorted Chromosome positions. The reason I need that is to describe my patterns  in a simpler way than by Chromosome start and stop.

Summary and Next Steps

• Going down the road of filling in patterns, I found something that I needed to correct in an earlier step
• This caused me to start afresh with cleaner tables.
• I was able to add a unique ordered Position ID to the combined V1/V2 table
• This position ID will be used to identify each Pattern for filling in missing alleles
• However, I have to re-do the patterns. This time, I will include the Position ID in the start and end of each pattern in my summary of Mom and Dad Patterns.

In my previous Blog, I looked at Whit Athey’s Principle 3 for my mom, my 4 siblings and myself. Based on that Priniciple and the previous 2, I phased our DNA up to a point. The next step in the phasing has to do with patterns.

Patterns

The patterns I am talking about are the patterns that the five siblings receive from either their mother or father.  For example, an AAAAB pattern means that the first 4 siblings received the same allele and the fifth sibling received a different one. I had mentioned previously that the patterns should be in this form:

• AAAAA
• AAAAB
• AAABA
• AAABB
• AABAA
• AABAB
• AABBB
• ABAAA
• ABAAB
• ABABB
• ABBBB

The first situation is a special case as this situation can happen within the other patterns ‘by accident’ as Athey puts it.

AAAAB Dad pattern

First I’ll look at a query to find an AAAAB pattern.

That Query results in this:

Except there are actually over 8,000 lines. I summarized the rough starts and stops in an Excel Spreadsheet:

This part can get a bit tedious. In Chromosome 2, I noted a possible break between 89 and 96M, so I’ll need to keep an eye on that. Highlighted in yellow are single patterns which may or may not be significant.

quality check

I took my AAAAB Query results and put them into an Excel spreadsheet. Then I subtracted the previous position number from the current position number to see where there were gaps. Then I filtered the gap to 1,000,000 or more positions:

This is my gap analysis. I highlighted the 7 million position gap where I put in an extra segment on the AAAAB pattern. This points out some of the single AAAAB patterns also.

mapping the initial results

Let’s look at Chromosome 13 between 28 and 87M. With an AAAAB pattern, that means that Joel, Sharon, Heidi and Jon match the same paternal grandparent. Lori matches a different one. However, we don’t know which paternal grandparent without a reference cousin. Fortunately, I have one. He is my dad’s first cousin. He would match on my paternal grandfather’s side. That grandfather is James Hartley, b. 1891:

Paternal cousin Jim matches the 5 siblings here:

As you may guess, Lori is on the bottom (#5). She has a crossover at about 85.5M according to Gedmatch. That means that before 85.5M she is matching on my father’s mother’s side: Marion Frazer. So, if I wanted to, I could start to map Chromosome 13. From 28 to 87M, I could say that 4 siblings got their DNA from their paternal Hartley grandfather and one sibling, Lori got hers from her paternal Frazer grandmother.

Further, I would expect an AAAAA pattern starting at 87M based on the gedmatch browser results above. The bad thing about an AAAAA pattern is that there is some missing DNA for the other grandparent. In this case, the Frazer DNA is lost on the right side of the map below. Another point is that these patterns change one letter at a time. So it makes sense that an AAAAB would go to an AAAAA. For example, an AAAAA would never go directly to an ABABA.

Here is a paternal only map of Chromosome 13 based on our very initial results:

aaaab Mom pattern

I notice that the formula that I used to find the AAAAB Dad pattern, I can move over to the mom side. So I might as well do this while I’m thinking of AAAAB pattenrs and put the results in Excel.

I randomly used Heidi as the ‘A’. So Lori not matching Heidi becomes the ‘B’. The results for this maternal query was much smaller with only 189 lines.

This was a lot easier. The Mom and Dad Patterns don’t interrelate with each other, so I have them on separate worksheets. Note that there is the same AAAAB pattern in the same starting place on Chromosome 13 as there was on the paternal side. This is a coincidence and the starting spot is a coincidence. This is just a rough number now and may be refined later. I could make a map of this also.

Here is a cousin on my mom’s father’s side:

Here she matches Joel, Sharon, Heidi and Lori from about 74-99M. Here is a map drawn on the Gedmatch browser and raw data phasing:

This shows what a AAAAB pattern looks like that is both paternal and maternal between 28 and 45M. I also show two crossovers for Lori: (Frazer to Hartley) and (Rathfelder to Lentz). In addition, Jon has to have a crossover from Rathfelder to Lentz and Lori has to have another crossover from Lentz to Rathfelder somewhere in the white spaces. There is a reason that I could tell the maternal A’s of the AAAAB pattern were Rathfelder even though our cousin match did not overlap that area. It is because the patterns do not change that fast as I explained above.

Now that I know which sibling has one of the paternal crossovers I can mark it on the Dad Pattern Spreadsheet:

I name the crossover column in the spreadsheet for the end of the pattern position, so it will be clear where it is. This is the ultimate goal of the process: to find the crossover locations and assign them to the siblings. Once this is done a map may be drawn for all the siblings.

The Next Step

In the next step, I could fill in the missing alleles between the Start and End positions of the AAAAB patterns. Here is how that will be done:

The highlighted row is where the AAAAB Pattern starts. Basically, what will happen is if there is at least one Allele in the first four positions, I will be able to fill in any of the other alleles in those first four positions with the same allele. However, in the last row, for example, there is just one G in the last position. We don’t know if the other four alleles will be a G or another letter. The row that has TTT??. We know that we can fill in the fourth T to TTTT?. However, the last allele we don’t know if it will be a T again or a different alelle. So we also need to leave that space blank.

However, I want to make sure I have all my patterns right, so I will look at all the patterns first and reconcile them.

AAABA Pattern

If I drew my map correctly above, I will be expecting Lori to have a maternal AAABA pattern on Chromosome 13. This should change to an AAABB pattern at about position 95M. I’m already on a maternal query, so I’ll start there.

I used Heidi again as the A. Now Jon is the B that is different than Heidi. I was surprised with the results as I only had this maternal pattern in Chromosome 1 and 23:

My prediction of a Chromosome 13 AAABA Pattern did not come true. I wonder what went wrong?

Paternal AAABA Pattern

Here is a partial summary of the Paternal AAABA Pattern:

On Chromosome 11, we see the AAABA pattern twice with an AAAAB pattern in-between. To go from an AAAAB to AAABA there has to be a transition pattern: either AAAAA or AAABB. Hopefully this prediction will be correct! That leads me to the AAABB pattern.

AAABB

This pattern requires a slight modification of my previous query:

This pattern is adjacent to the AAABA Pattern, so I will be able to assign some crossovers:

These crossovers belong to Jon and Lori as Jon is in the next to the last position of the patterns and Lori is in the last position of the patterns. Note that in Chromosome 19, Lori goes from an AAABA to an AAABB at about 5M. However, there is a rogue AAABB in the AAABA pattern at around 3M. That could be due to a misread or a mutation. I’m not sure. Jon has a crossover on Chromosome 8. These are all Lori and Jon crossovers, due to the positions of the pattern changes we are looking at. The changes are all in the last two positions.

AAABB Maternal

I’m still getting very few crossovers here. I’m not sure why:

I’m not sure why the maternal side is not keeping up with the paternal. I have no crossovers here yet.

AABAA

Following my alphabetical reasoning, AABAA is my next pattern. I’ll start with the maternal:

I changed to having my [Joel’s] allele the ‘A’ in the Pattern. The results look right:

It seemed like there was a break in Chromosome 5 between 46 and 50M.

AABAA Paternal

On Chromosome 5, there was a gap similar to the one on the Maternal side.

Centromeres

According to ISOGG, these are the Build 37 Centromeres:

This is good information to have. I assume that the Centromere is not counted, so I will ignore the Chromosome 5 missing area and make a note that the centromere is there. This also makes a difference on all the results.

AABAB: Are We There Yet?

Here are Heidi’s first crossovers. I’ve also heard of crossovers referred to as cut points. I am noting where the centromere is – though not quite spelling it correctly above.

Here is the Maternal AABAB. I am still annoyed that there are so few patterns. They seem to be missing for some reason:

I suppose, if this trend continues, I could do the project over and add in my mother’s and my FTDNA raw DNA results.

AABBA

I didn’t find any AABBA Patterns on the maternal side. However, that was with a query using my results as A. However, from my previous Blog, I recall this chart:

This showed that on the Mom side, Jon and Lori had the most alleles. I’ll run the query again this way:

Still no patterns.

Here are some more Dad Patterns:

However, there are a few problems. Chromosome 17 is missing a pattern. I can solve this by looking at the original table.

Here the pattern is AABAA.

The next problem is that there are two patterns in one spot on Chromosome 22. I ran pattern AAABA again and see it should have ended earlier:

Here is the right answer below that also shows a Heidi crossover at that location:

AABBB

Paternal

Maternal side

Still nothing.

ABAAA

Maternal side

Paternal

This side had more patterns.

ABAAB

Had several of ABAAB Patterns on the dad side, but only one on the mom side. I think that there is a fill-in step that fills in the mom side from the dad side that may correct this later.

ABABA, ABABB, ABBAA

I did notice a Dad Pattern discrepancy on Chromosome 6:

There are three single patterns, I figured were discrepancies. However, there appeared to be a longer AAABB Pattern within the ABBAA Pattern. This is where it helps to look at the raw data.

The blue section is the start of the AAABB rogue pattern that I had. However, a closer examination reveals that this pattern is not continuous from position 30514810 to 30594827. Between those two there are a lot of ABBAA patterns. This is clear at position 30544401. However, this is also clear wherever the first 2 alleles are different. For example, on the last line, I see GT???. This will be filled in with GTTAA as this is within the ABBAA Pattern area. So what happened was that there were two single AAABB patterns. When I did the query for these, it looked like the pattern was continuous, but it was not. Based on the above, I’ve modified my Dad Pattern Spreadsheet to show two single discrepancies:

I won’t overwrite this information, but I will keep it in mind for later in case it is important. If this was a real crossover, it would be mine. However, crossovers in the middle of a chromosome don’t change that fast for one person on one copy of their chromosome.

Some of the Dad Pattern Crossovers are starting to fill in:

Starts and Ends of Chromosomes

At some point, it is important to know where the Chromosomes start and end. The testing companies don’t always start at the beginning positions of each chromosome. The ends are different also based on the lengths of the chromosomes.

I was able to find what I was looking for using a min/max Query in Access. I took my table with the 900,000 plus alleles and made a query that looks like this:

When I run the Query, I get this helpful table:

This tells me the start and end locations for each of the chromosomes that I am looking at.

I put this into Excel and highlighted the information in purple. Then I sorted it into my mom and dad pattern spreadsheets:

Now, I can tell that I am near the beginning and the end of Chromosome 20 with the pattern locations. However, on Chromosomes 21 and 22, there is still room for more at the beginning of those Chromosomes. As the Chromosome 20 patterns are complete, this also tells me that my sister Lori has no paternal crossover on Chromosome 20.

ABBAB, ABBBA, and ABBBB

These are the last three patterns, not counting AAAAA. I finally have one crossover on the maternal side. It is on the X Chromosome:

I have a mess to clean up on Paternal Chromosome 2 :

There appear to be two patterns occupying the same space between 123 and 128M, which is not good. I’ll take a look at my Table: It appears that the AAAAB at 127,841,390 is a one-time occurrence. Here is my correction:

Note that there is still a gap at AAAAB. There may be an AAAAA Pattern stuck in there.

Lessons Learned and Next Up

• It is good to document the process in case something goes wrong
• The start and end points are needed for each chromosome
• The start and end for each centromere is needed also
• Attention is needed for the location of each crossover and who it goes to as this is a main point of all the work.
• Changes along each copy of the chromosome are gradual. They happen one at a time and those one at a time changes correspond to siblings.
• Next up is filling in the blanks. That was discussed briefly in this Blog.

In my previous Blog, I started to phase 5 siblings based on their raw data and the raw DNA data from their mom. I looked at homozygous results. That is, when each sibling had the same allele, it meant that they got one of each of those same alleles from each parent. Also when my Mom had homozygous results, say GG, she had to have given one of those G’s to each of her children in that location.

I am using an Athey paper on Phasing from 2010. I looked at his first 2 principles in my previous Blog. Here is Principle 3:

5 Phased Sibs Update: V1 and V2

I did all my Principle 3 phasing and here is the update:

What is a little surprising is that Jon and Lori who were tested as AncestryDNA V2 had more Mom-phased alleles. I did mention above that they were getting extra phasing on SNPs that they hadn’t tested from their mom, but I didn’t realize how much.

I mentioned in my previous blog that the combined number of SNPs tested between V1 and V2 is 942,269. That number represents the merging of V1 and V2.

Also some of the specifics are a bit off. For example, my numbers include phased results for myself from my dad (16,536) on the X Chromosome. Well, I didn’t get an X from my dad. This means that the JoelfromDad and JonfromDad numbers above are a bit high.

Next up: DNA patterns

In many past Blogs, I have written about the raw DNA data of my siblings and my mother. They can be searched in my Blogs under “Raw DNA Data”. First, I looked at three siblings compared with our mom. Next I looked at the results of four siblings. Now I have a 5th sibling tested.

Phasing With Raw DNA Data

The reference I use is Whit Athey’s 2010 paper called, Phasing the Chromosomes of a Family Group When One Parent is Missing.

Basically, Whit uses certain rules and iterative processes to fill in blanks of what the parent’s alleles would be as passed down to their children. From these lists of alleles one can see patterns. From the patterns, one can see where the maternal and paternal crossovers would be. The process is similar to the visual phasing process developed by Kathy Johnston. However, Kathy’s version does not require the use of a parent. Also Kathy’s version does not require looking at hundreds of thousands of alleles.

Lori’s raw data

The fifth sibling in my family to have an autosomal DNA test is Lori. She tested at Ancestry DNA. First I unzipped her results. They open in notepad. I then opened those results in Excel, so all the data would align in columns.

The columns are a SNP ID, the Chromosome, the position on the Chromosome in Build 37 and allele 1 and 2. I added Lori to the allele columns, so I could distinguish between siblings when comparing. One quirk is that when I convert from text to Excel, the blanks in the allele columns go to zeros. I then have to search for all zero’s in those columns and replace them with blanks. The blanks are no-calls.

This data shows Lori’s alleles unsorted. We do know that where she has a C and a C, that one is from her dad and one from her mom. However, where she has an A and G on Line 380436, we don’t know which is from her dad and which is from her mom.

Lori’s DNA in MS Access

I didn’t realize I could upgrade my old computer to Windows 10. It was just new enough to do that. When I did that, I rented Office 365 which includes Access. Access is good for comparing large amounts of data. Lori has 666,531 lines of data. There are 2 alleles for each position. So with six sets of data, that is a lot of alleles. I figure about 8 million. However, the crossovers occur at a distinct point. Finding crossovers is like finding a needle in a haystack.

First I import Lori’s Excel File into Access. It looks pretty much the same there. Except that Access adds an extra ID to keep track of things. Next I want to make an Access Query based on Athey’s Principal 1:

Principle 1—If a person is “homozygous” at a location —that is, having the same base on each of the two chromosomes of a pair, then obviously at that location it is possible to know with certainty that both chromosomes of the pair have that base at that location, but this is an almost trivial form of phasing.

Principle 1 in Access Query form

Here is Lori’s first query in design view:

It’s a bit small. All I did is put all of Lori’s imported raw data into a query. Then in the last columns I created a field called Lorifromdad. Then there is a formula that says if Lori’s allele1 is the same as her allele2, then put in allele 2. When I run that query, I get this:

Next, I want Lori from mom, which will look the same as Lori from dad. This is easy. I can just copy the same formula and give it a different name:

Also, I forgot that Ancestry has other DNA information in the raw data that I don’t need so I need to restrict the data to Chromosomes 1-23:

It’s nice to check the results to make sure you are getting what you want. This looks pretty simple, but Access does this operation over 600,000 times, so it saves a lot of time.

Next I add Jon:

I have the same kind of formula to Jon’s homozygous results from his mom and dad. I made an equal join in the query above. Note that Jon and Lori both tested with AncestryDNA V2. That means that they have the same SNPs tested. My 2 sisters, my mom and I all tested with V1. So we have to be careful with these joins. If I was to have used an equal join between a V1 and a V2 test, I would only get the results which were common to both.

When I view the query above, it looks like this:

Note that on the third line, Lori has homozyzous results and Jon does not.

Adding AncestryDNA V1 and V2 raw results

The next step is that I would like to carefully add the V1 homozgous results to the V2 homozygous results. Also I would like to make a large table out of what I get.

• On my existing V1 Homozygous Table, I have 700,153 rows
• On my new V2 Homozygous Table, I have 666,153 rows
• The V1 SNPs that are the same as the V2 SNPs are 424037 rows or results

That means that I would like to have a table that has the V1 results plus the V2 results, minus the results in common, so that should be 942269 rows. Somehow I ended up with such a table. I know that’s not very scientifically reproducible, but that is what happened. I’m not sure how important it is to have the V2 results as they won’t phase with my mom. However, I’ll have them in case I need them.

The results of the query left the two V2 siblings’ results on the bottom of the table, but they can easily be sorted:

Principle 2

According to the Athey paper:

This principle lends itself to Access. Basically, I want to tell Access that if Mom has the same two alleles, then show that each child got that allele from her. However, there are a few considerations. If mom has no-reads and the child doesn’t, then we don’t want to overwrite a good read with a bad one. The other consideration is, if mom had an incorrect read and the children had a correct one, we wouldn’t want to overwrite that either. However, I don’t know how rare that is. I guess it is pretty rare. I did a query to check and didn’t find any such instance. So that is one less thing to worry about.

Principle 2 in access

I want to say if mom has two non-null alleles that are the same, put that allele in as from her for all her children. Looking at my old queries, it looks like I need an Update Query. First, I copy my previous table of results to a new table called tbl5sibsMomHomozygous. I’ll try this query:

Before I update, I’ll take a view:

If I take out the ‘And is not null’ statement, I get the same results. I then changed the syntax to ‘Is not Null’ first and got one less record: 481977. It makes me wonder what that record is? I’ll use my second wording as it may be more accurate. Next I hit the !Run button and it updates the table I recently made.

This will give V1 mom alleles to Lori and Jon even when they weren’t tested for them. Here is an updated table view of just the alleles from Dad and the alleles from Mom:

I picked these results at random about halfway down the table. It looks like about half the alleles are filled in already. So now my siblings are more than half phased. The first 5 rows are alleles from Dad for each of the 5 siblings. The 2nd five rows are alleles from Mom for those same siblings.

a pattern preview

The highlighted row shows a pattern from Dad and one from Mom. The first row also shows the same pattern. This is what we will be looking at later in more detail to determine crossovers on the maternal and paternal side. This is what I’ll call the ABABA pattern for both. Here it is coincidental that both the Dad and Mom patterns are the same. Obviously with 5 siblings, there will be a lot of different types of patterns:

• AAAAA
• AAAAB
• AAABA
• AAABB
• AABAA
• AABAB
• AABBB
• ABAAA
• ABAAB
• ABABB
• ABBBB

Those are the combinations that I can think of right now. AAAAA is a special case. This could mean that all five siblings could share the same grandparent or sometimes an AAAAA pattern is that way by coincidence. Where the maternal or paternal pattern changes is where the crossover is. This pattern should be gradual. That is, only one letter should change in a pattern change. For example, ABABA may change to ABABB or AAABB. There are many possibilities but only one letter will change. The placement of the letter represents one sibling. So that sibling will own that crossover. For example, a maternal ABABA to ABABB change would represent a maternal crossover for Lori as she is in the last position on my table. The place where the A goes to a B is the location of the crossover.

Next Up

Next up is Athey’s Principal 3 as it applies to 5 siblings and a Mom.