How are raw well numbers determined?

[seqAll]

...Basically, about 1/2 the enriched beads loaded are dud beads...
Phillip

Thanks Philip, that clears my confusion.

[pmiguel]

Hi Phillip and seqAll,

[...] the software manual Part B pp. 12-13 describes the series of signal corrections and filters that are applied to yield Raw Wells.
[...] In our experience, the overwhelming majority of beads loaded do end up in wells of the PicoTiterPlate. A small percentage end up in non-desired areas (like the gasket), and as you point out, a small percentage remain in the supernatant. However, some beads that are deposited will not sequence well for a variety of reasons - i.e. insufficient number of copies of template, null beads from lower stringency of enrichment, etc. Wells containing beads of that nature will not yield a Raw Well.

Hi Jason,
It is great to see an answer to this mystifying process. What you write more or less confirms my suspicions. Basically, about 1/2 the enriched beads loaded are dud beads--producing little or no signal or displaying some other sign that makes them quickly discountable?
That said, additional clarity would be welcome.
Specifically, a break down of which filters were invoked to exclude each non-raw well for a given run would be very helpful in assessing input library issues. (Note, that the only "filter" mentioned in pp 11-12 of the SW-Manual_PartB is a "ghost well" filter. And the bullet list at the top of page 10 makes it ambiguous as to whether this filter is even invoked in non-Amplicon pipelines.)
Is this information recorded anywhere? I can parse log files, if necessary.
Thanks,
Phillip

[Zaag]

We had 2.273.997 raw wells in the last run with 2 regions.

[454Sequencing]

Hi seqAll,

Unless I am mistaken you might be confusing Raw Wells with quality filtered reads. Quality filtered reads are themselves a percentage of Raw Wells, since they represent the output of additional quality filtering processes. For example, if you had 2M Raw Wells total and of those 65% pass quality filtering, the resulting yield for the run is 2M * 65% = 1.3M quality filtered reads.

You can find more details on quality filtering in Part B of the software manual, Section 1.4.2, starting on pp. 13. For more information I'd recommend following up with your local Roche representative as they likely can provide some additional training if desired.

Best regards,
Jason

[seqAll]

Thanks Jason,

One question I don't understand. Most of the time we have the filtered% around 50% to 70%. If the overwhelming majority of beads (2 M per region) loaded do end up in wells of the PicoTiterPlate, there will be about 1 M to 1.4 M filtered reads, that is 2 M to 2.8 M high quliaty reads per run. But we know that the yield per run is about 1 M.

Don't know where I got wrong.

[454Sequencing]

Hi Phillip and seqAll,

My first recommendation would be to consult the 454 software manual for more information on what defines Raw Wells. The Manuals section of the my454 customer site (http://www.454.com/my454) has recently been updated with the current (v2.3) software manual. Specifically, the software manual Part B pp. 12-13 describes the series of signal corrections and filters that are applied to yield Raw Wells.

As for the topic of bead loading & how it relates to Raw Wells - We have optimized the bead loading recommendations (for example, 2M beads per region of a 2-region gasket) to maximize the number of High Quality Reads while minimizing the number of beads loaded. As part of the same efficiency measures we also optimized the emPCR enrichment procedures to yield the most, high quality, sequence-able beads.

In our experience, the overwhelming majority of beads loaded do end up in wells of the PicoTiterPlate. A small percentage end up in non-desired areas (like the gasket), and as you point out, a small percentage remain in the supernatant. However, some beads that are deposited will not sequence well for a variety of reasons - i.e. insufficient number of copies of template, null beads from lower stringency of enrichment, etc. Wells containing beads of that nature will not yield a Raw Well.

All that said, if you are seeing any change in performance I would certainly recommend following up with your local Roche representative to aid in troubleshooting.

Best regards,
Jason

[seqAll]

I see that the numbers of beads loaded and pulled off do not adds up to the original total.

Is the variation of the bead counter big or small? Maybe I was too warry about the accuracy of bead counter.

The volume of well is designed to take one single bead + some other small beads such as enzyme beads etc. Could there be 1.5 beads in one well? - A half in the well, another half above the surface. This might connect two good wells to become bad. Maybe this is not a problem after centrifuge. I don't know.

[pmiguel]

Interesting question! Would like to know the answer as well.

It seems like the way of beads deposition has a large space to optimise. Too dense beads, say if you can fill all wells with beads, leads to totally unreadable signal. Too loose beads wastes the reagents.

For the entire plate, one may control the average bead density. Locally, there are always some too dense, some too loose.

Antoher post in the forum (can't remenber now) which shows the plots of untrimmed vs trimmed reads seems to say that too dense beads also lead to short read.

Yes, the 454 manuals warn of overloading the PTP.

I have had several people tell me that only some of the beads added to the PTP actually end up in a well. The implication being that ~50% are pulled off the plate during steps subsequent to deposition. However, when we count the beads that do get pulled off, we can account for around 10-15% of the total.

The titanium PTP is said to have 3.6 million wells. The protocol calls for loading 4 million beads (2 million/region). The number of beads recovered during steps subsequent to sample bead loading is consistent will nearly all the beads finding a well. Yet, raw well counts tend to be around 2 million for the entire PTP. On the SOLiD we would call that a 50% "deposition efficiency". Except, by the process of elimination, we know that nearly all the beads (~80-90%) are actually being deposited into a well.

Recently this "efficiency" has been falling for us. While the results have been acceptable (around 400 megabases of sequence), both sample and control bead keypass numbers are 35% of the total loaded. (With keypass numbers >90% of raw well numbers).

--
Phillip

[seqAll]

Interesting question! Would like to know the answer as well.

It seems like the way of beads deposition has a large space to optimise. Too dense beads, say if you can fill all wells with beads, leads to totally unreadable signal. Too loose beads wastes the reagents.

For the entire plate, one may control the average bead density. Locally, there are always some too dense, some too loose.

Antoher post in the forum (can't remenber now) which shows the plots of untrimmed vs trimmed reads seems to say that too dense beads also lead to short read.

[pmiguel]

What constitutes a "raw well" for a Titanium run? "keypass" wells require the 4 key bases. But how is a raw well identified?

Typically 4 million beads are loaded and, if all goes well, you see about 2 million raw wells. What happens to the other 2 million beads?

The Titanium Chemistry overview says there are 3.6 million wells per PTP. So there are wells available for most of the beads. Not many sample beads appear to be pulled off during deposition. When we count the number of sample beads in the pooled liquid removed during deposition, there are about 10% of the sample beads there.


--
Phillip