I believe, your signal is about 4mV amplitude which is sumperimposed over 126mV DC, right? But, the frequency, you mean the repetition rate of incident light pulse?
You are using a transresistance amplifier to convert photo diode current to output voltage, here, your gain is set by the feedback resistor. So, definitely you can expect huge amount of thermal noise. Also, even at normal lab environment itself, the noise floor would be as high as 40mV (switching of flurescent light etc.). Hence, you must amplify the signal at least as high as 80 to 100mV. So that you can set a marginal level of hysterisis to the final stage comparator. For the DC, at the final stage of amplification, before feeding to the comparator, provide a RC high pass filter, it will just kill the DC. I shall list..
1. Provide high gain stages, may be two stages to make the minimum amplitude at least 80mV
2. Kill the DC
3. Set reference to 35mV with hysterisis +/-20mV to the final comparator